{ "pages": { "01.AI": { "title": "01.AI", "type": "organization", "words": 535, "refs": 5, "outbound": [ "Baichuan", "Chinese_AI_labs", "DeepSeek", "DeepSeek-R1", "Hugging_Face", "Llama_(model_family)", "MMLU", "Moonshot_AI", "Pretraining", "Qwen_team", "Scaling_laws", "Yi_(model_family)" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "AFM-4.5B": { "title": "AFM-4.5B", "type": "model", "words": 382, "refs": 5, "outbound": [ "Arcee_AI", "Hugging_Face", "Instruction_tuning", "Knowledge_distillation", "Large_language_model", "Llama_3.1", "Mixture_of_experts", "Pretraining", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Mini", "Trinity_Nano" ], "categories": [ "Arcee models", "Models", "2025 model releases", "Open-weight models" ] }, "AI21_Labs": { "title": "AI21 Labs", "type": "organization", "words": 1283, "refs": 7, "outbound": [ "Aleph_Alpha", "Amazon_Titan", "Anthropic", "Claude_3.5_Sonnet", "Cohere", "GPT-2", "GPT-3", "GPT-4o", "Google_DeepMind", "Hugging_Face", "Hybrid_architecture_(LLM)", "Instruction_tuning", "Jamba", "Jurassic_(models)", "Large_language_model", "Megatron-Turing_NLG", "MiniMax", "Mistral_AI", "Mixture_of_experts", "NVIDIA", "Nemotron", "OpenAI", "State-space_model", "Stated_missions_of_AI_labs", "Technology_Innovation_Institute", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] }, "AI_2027": { "title": "AI 2027", "type": "essay", "words": 802, "refs": 8, "outbound": [ "Chain-of-thought_prompting", "Large_language_model", "Machines_of_Loving_Grace", "Mechanistic_interpretability", "OpenAI", "Pretraining", "Reinforcement_learning_with_verifiable_rewards", "Seminal_AI_essays", "Simulators_(essay)", "Situational_Awareness_(essay)", "Sparse_autoencoder", "Test-time_compute", "The_Most_Important_Century", "What_Failure_Looks_Like", "o1" ], "categories": [ "Essays", "AI governance" ] }, "AI_and_Compute": { "title": "AI and Compute", "type": "essay", "words": 594, "refs": 5, "outbound": [ "AlexNet", "AlphaZero", "GPT-3", "Ilya_Sutskever", "NVIDIA", "OpenAI", "Scaling_laws", "Seminal_AI_essays", "Situational_Awareness_(essay)", "The_Bitter_Lesson", "The_Scaling_Hypothesis" ], "categories": [ "Essays", "Hardware and compute" ] }, "ALBERT": { "title": "ALBERT", "type": "model", "words": 382, "refs": 3, "outbound": [ "BERT", "DistilBERT", "ELECTRA", "GPT-2", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "Masked_language_modeling", "Pretraining", "RoBERTa", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2019 model releases", "Open-weight models" ] }, "Activation_steering": { "title": "Activation steering", "type": "technique", "words": 1827, "refs": 10, "outbound": [ "Activation_verbalizer", "Anthropic", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_Opus_4.6", "Constitutional_AI", "Eliciting_Latent_Knowledge", "Feed-forward_network", "GPT-2", "GPT-J", "Gemma", "LLaMA", "Large_language_model", "Llama_2", "Logit_lens", "Mechanistic_interpretability", "Reinforcement_learning_from_human_feedback", "Self-attention", "Sparse_autoencoder", "Superposition_(interpretability)", "Transformer_(architecture)" ], "categories": [ "Concepts", "Alignment and safety", "Inference" ] }, "Activation_verbalizer": { "title": "Activation verbalizer", "type": "technique", "words": 1810, "refs": 9, "outbound": [ "Activation_steering", "Anthropic", "Chain-of-thought_prompting", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_Opus_4.6", "Eliciting_Latent_Knowledge", "GPT-2", "Instruction_tuning", "Large_language_model", "Logit_lens", "Mechanistic_interpretability", "Reinforcement_learning_from_human_feedback", "Sparse_autoencoder", "Superposition_(interpretability)", "Transformer_(architecture)" ], "categories": [ "Alignment and safety", "Concepts" ] }, "Adam_(optimizer)": { "title": "Adam (optimizer)", "type": "concept", "words": 1284, "refs": 7, "outbound": [ "BERT", "Backpropagation", "Cross-entropy_loss", "Distributed_training", "GPT-3", "Geoffrey_Hinton", "Large_language_model", "Pretraining", "T5", "Transformer_(architecture)" ], "categories": [ "Concepts", "Training methods" ] }, "Ai2": { "title": "Ai2", "type": "organization", "words": 527, "refs": 5, "outbound": [ "Anthropic", "CLIP", "Direct_preference_optimization", "EleutherAI", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Llama_3.1", "MMLU", "Meta_AI", "Mistral_AI", "Molmo", "OLMo", "OpenAI", "Pretraining", "Pythia", "Qwen2", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Tulu_3" ], "categories": [ "Organizations", "Open-source AI" ] }, "Aleph_Alpha": { "title": "Aleph Alpha", "type": "organization", "words": 944, "refs": 6, "outbound": [ "Anthropic", "BLOOM", "Cohere", "Falcon_(models)", "GPT-3", "GPT-4", "Hugging_Face", "Large_language_model", "Mistral_7B", "Mistral_AI", "Mixtral_8x7B", "OpenAI", "Pretraining", "Scaling_laws", "Stated_missions_of_AI_labs", "Technology_Innovation_Institute", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] }, "AlexNet": { "title": "AlexNet", "type": "model", "words": 1074, "refs": 9, "outbound": [ "Alex_Krizhevsky", "AlphaGo", "Backpropagation", "Boltzmann_machine", "Dropout", "Geoffrey_Hinton", "Ilya_Sutskever", "ImageNet", "NVIDIA", "OpenAI", "Safe_Superintelligence", "Seq2seq", "Softmax", "Vision_Transformer", "Yoshua_Bengio" ], "categories": [ "Models", "Pre-transformer systems" ] }, "Alex_Krizhevsky": { "title": "Alex Krizhevsky", "type": "person", "words": 1041, "refs": 10, "outbound": [ "AlexNet", "BERT", "Backpropagation", "Boltzmann_machine", "Distributed_training", "Dropout", "GPT_(model_family)", "Geoffrey_Hinton", "Google_DeepMind", "Ilya_Sutskever", "ImageNet", "Large_language_model", "NVIDIA", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "People", "Machine learning researchers" ] }, "Alpaca": { "title": "Alpaca", "type": "model", "words": 402, "refs": 4, "outbound": [ "Databricks", "GPT-2", "GPT-3.5", "GPT-J", "Hugging_Face", "Knowledge_distillation", "LLaMA", "Llama_(model_family)", "MMLU", "Meta_AI", "OpenAI", "Pretraining", "Supervised_fine-tuning", "Transformer_(architecture)", "Vicuna" ], "categories": [ "Models", "Open-weight models", "2023 model releases" ] }, "AlphaFold": { "title": "AlphaFold", "type": "model", "words": 1419, "refs": 12, "outbound": [ "AlexNet", "AlphaGo", "AlphaZero", "Backpropagation", "Cross-entropy_loss", "Demis_Hassabis", "Google_DeepMind", "ImageNet", "John_Jumper", "Large_language_model", "Masked_language_modeling", "Meta_AI", "Multi-head_attention", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Models", "Google models" ] }, "AlphaGo": { "title": "AlphaGo", "type": "model", "words": 1393, "refs": 10, "outbound": [ "AlexNet", "AlphaFold", "AlphaZero", "Arthur_Samuel", "David_Silver", "DeepSeek-R1", "Demis_Hassabis", "Google_DeepMind", "Ilya_Sutskever", "Large_language_model", "Reinforcement_learning_with_verifiable_rewards", "Richard_Sutton", "TD-Gammon", "Test-time_compute", "o1" ], "categories": [ "Models", "Google models", "Pre-transformer systems" ] }, "AlphaZero": { "title": "AlphaZero", "type": "model", "words": 495, "refs": 4, "outbound": [ "AlphaGo", "Arthur_Samuel", "David_Silver", "DeepSeek-R1", "Demis_Hassabis", "Google_DeepMind", "Large_language_model", "Reinforcement_learning_with_verifiable_rewards", "TD-Gammon", "Test-time_compute", "o1" ], "categories": [ "Models", "Google models", "Pre-transformer systems" ] }, "Amazon_Nova": { "title": "Amazon Nova", "type": "model", "words": 341, "refs": 3, "outbound": [ "Amazon_Titan", "Anthropic", "GPT-4o", "Google_DeepMind", "Knowledge_distillation", "MMLU", "NVIDIA", "OpenAI", "Transformer_(architecture)" ], "categories": [ "Models", "Multimodal models", "2024 model releases" ] }, "Amazon_Titan": { "title": "Amazon Titan", "type": "model", "words": 397, "refs": 3, "outbound": [ "Amazon_Nova", "Anthropic", "Claude_(model_family)", "Command_(models)", "Google_DeepMind", "Jurassic_(models)", "Large_language_model", "Llama_2", "Stability_AI", "Stable_Diffusion", "SynthID", "Transformer_(architecture)" ], "categories": [ "Models", "2023 model releases", "Multimodal models" ] }, "Analytic_distillation": { "title": "Analytic distillation", "type": "concept", "words": 417, "refs": 5, "outbound": [ "BERT", "Compositional_generalization", "DistilBERT", "GPT-4", "In-context_learning", "Knowledge_distillation", "Mechanistic_interpretability", "Sparse_autoencoder" ], "categories": [ "Concepts", "Training methods" ] }, "Andrej_Karpathy": { "title": "Andrej Karpathy", "type": "person", "words": 1787, "refs": 14, "outbound": [ "Adam_(optimizer)", "AlexNet", "Anthropic", "Backpropagation", "CLIP", "Codex_(2021_model)", "Distributed_training", "Dropout", "GPT-2", "GPT-3", "GPT-4", "Ilya_Sutskever", "ImageNet", "In-context_learning", "InstructGPT", "LSTM", "Large_language_model", "Next-token_prediction", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Reward_model", "Self-attention", "Seq2seq", "Software_2.0", "Supervised_fine-tuning", "Tokenization", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "People", "Machine learning researchers" ] }, "Andrew_Ng": { "title": "Andrew Ng", "type": "person", "words": 1195, "refs": 9, "outbound": [ "AlexNet", "Backpropagation", "Baidu", "Google_DeepMind", "Hugging_Face", "Ian_Goodfellow", "Ilya_Sutskever", "Large_language_model", "NVIDIA", "OpenAI" ], "categories": [ "People", "Machine learning researchers" ] }, "Ant_Group": { "title": "Ant Group", "type": "organization", "words": 515, "refs": 5, "outbound": [ "Chinese_AI_labs", "DeepSeek", "DeepSeek-V3", "Huawei", "Hugging_Face", "Ling_(models)", "LongCat", "Mixture_of_experts", "Moonshot_AI", "NVIDIA", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "iFlytek" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Anthropic": { "title": "Anthropic", "type": "organization", "words": 654, "refs": 6, "outbound": [ "Claude_(model_family)", "Claude_2", "Claude_3", "Claude_4", "Claude_Fable_5", "Claude_Sonnet_5", "Constitutional_AI", "GPT-4", "Google_DeepMind", "Large_language_model", "MMLU", "Machines_of_Loving_Grace", "Mechanistic_interpretability", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] }, "Apple_foundation_models": { "title": "Apple foundation models", "type": "model", "words": 426, "refs": 5, "outbound": [ "AFM-4.5B", "Arcee_AI", "GPT-4o", "Gemini_(model_family)", "Instruction_tuning", "Large_language_model", "Mixture_of_experts", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Models", "2024 model releases", "Multimodal models" ] }, "Arcee_AI": { "title": "Arcee AI", "type": "organization", "words": 1290, "refs": 9, "outbound": [ "AFM-4.5B", "Claude_Opus_4.6", "DeepSeek", "DeepSeek-R1", "Distributed_training", "GLM-5", "Hugging_Face", "Instruction_tuning", "Kimi_K3", "Knowledge_distillation", "LLaMA", "Large_language_model", "Llama_3", "Llama_3.1", "Llama_4", "MMLU", "Meta_AI", "Mistral_AI", "Mixture_of_experts", "Moonshot_AI", "NVIDIA", "Pretraining", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Supervised_fine-tuning", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Large_Thinking", "Trinity_Mini", "Trinity_Nano", "Zhipu_AI" ], "categories": [ "Organizations", "Frontier labs" ] }, "Arthur_Samuel": { "title": "Arthur Samuel", "type": "person", "words": 1246, "refs": 12, "outbound": [ "AlphaGo", "AlphaZero", "Backpropagation", "David_Rumelhart", "David_Silver", "Feed-forward_network", "Frank_Rosenblatt", "Geoffrey_Hinton", "Google_DeepMind", "Large_language_model", "NETtalk", "Paul_Werbos", "Perceptron", "Richard_Sutton", "TD-Gammon", "Transformer_(architecture)" ], "categories": [ "People", "Computer science pioneers" ] }, "Attention_sink": { "title": "Attention sink", "type": "concept", "words": 1579, "refs": 5, "outbound": [ "Large_language_model", "Llama_2", "OpenAI", "PagedAttention", "Pretraining", "Pythia", "Self-attention", "Sliding-window_attention", "Softmax", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "Architectures", "Concepts", "Inference" ] }, "Aya": { "title": "Aya", "type": "model", "words": 354, "refs": 4, "outbound": [ "BLOOM", "Cohere", "Command_(models)", "Gemma", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "Mistral_7B", "T5", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2024 model releases", "Multimodal models" ] }, "BERT": { "title": "BERT", "type": "model", "words": 324, "refs": 3, "outbound": [ "ALBERT", "DistilBERT", "ELECTRA", "Embedding_(machine_learning)", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Large_language_model", "MMLU", "Masked_language_modeling", "OpenAI", "Pretraining", "RoBERTa", "Self-attention", "Transformer_(architecture)", "Word2vec" ], "categories": [ "Google models", "Models", "2018 model releases" ] }, "BLIP": { "title": "BLIP", "type": "model", "words": 388, "refs": 3, "outbound": [ "BERT", "CLIP", "Flamingo", "Flan-T5", "Hugging_Face", "Instruction_tuning", "LLaVA", "Large_language_model", "OPT", "Transformer_(architecture)", "Vicuna", "Vision_Transformer" ], "categories": [ "Multimodal models", "Open-weight models", "Models", "2022 model releases", "2023 model releases", "Open-source AI" ] }, "BLOOM": { "title": "BLOOM", "type": "model", "words": 377, "refs": 3, "outbound": [ "Chinchilla", "Distributed_training", "EleutherAI", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Gopher", "Hugging_Face", "LLaMA", "Large_language_model", "Layer_normalization", "Megatron-LM", "Meta_AI", "OLMo", "OPT", "Pythia", "Transformer_(architecture)", "YaLM-100B", "Yandex" ], "categories": [ "Models", "Open-weight models", "Open-source AI", "2022 model releases" ] }, "Backpropagation": { "title": "Backpropagation", "type": "technique", "words": 2032, "refs": 22, "outbound": [ "Adam_(optimizer)", "Cross-entropy_loss", "Distributed_training", "Feed-forward_network", "Geoffrey_Hinton", "LSTM", "Large_language_model", "NETtalk", "Paul_Werbos", "Perceptron", "Pretraining", "Scaling_laws", "Seq2seq", "Transformer_(architecture)" ], "categories": [ "Training methods", "Concepts" ] }, "Baichuan": { "title": "Baichuan", "type": "organization", "words": 594, "refs": 5, "outbound": [ "01.AI", "Chinese_AI_labs", "DeepSeek", "Hugging_Face", "Llama_(model_family)", "MMLU", "MiniMax", "Moonshot_AI", "Qwen_(model_family)", "StepFun", "Zhipu_AI" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Baidu": { "title": "Baidu", "type": "organization", "words": 572, "refs": 5, "outbound": [ "Andrew_Ng", "Chinese_AI_labs", "DeepSeek", "DeepSeek_(model_family)", "Doubao", "ERNIE_5", "ERNIE_Bot", "GPT-4", "Hugging_Face", "Large_language_model", "MMLU", "NVIDIA", "Pretraining", "Qwen_(model_family)", "Qwen_team", "Tencent" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Boltzmann_machine": { "title": "Boltzmann machine", "type": "concept", "words": 1153, "refs": 10, "outbound": [ "AlexNet", "Backpropagation", "Dropout", "Geoffrey_Hinton", "Hopfield_network", "John_Hopfield", "NETtalk", "Perceptron", "Pretraining", "Softmax" ], "categories": [ "Concepts", "Architectures" ] }, "ByteDance": { "title": "ByteDance", "type": "organization", "words": 545, "refs": 5, "outbound": [ "Anthropic", "Chinese_AI_labs", "DeepSeek", "Doubao", "Hunyuan", "MMLU", "NVIDIA", "OpenAI", "Qwen_team", "Seed-OSS", "Seedream", "Tencent" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "CLIP": { "title": "CLIP", "type": "model", "words": 403, "refs": 4, "outbound": [ "DALL-E", "DALL-E_2", "FLUX", "Flamingo", "GPT-3", "HunyuanVideo", "ImageNet", "Molmo", "OpenAI", "Self-attention", "Stable_Diffusion", "Transformer_(architecture)", "Vision_Transformer", "Whisper" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "Open-weight models", "2021 model releases" ] }, "Cerebras-GPT": { "title": "Cerebras-GPT", "type": "model", "words": 294, "refs": 3, "outbound": [ "Cerebras_Systems", "Chinchilla", "Distributed_training", "EleutherAI", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Hugging_Face", "NVIDIA", "Nemotron", "Pythia", "Scaling_laws", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2023 model releases" ] }, "Cerebras_Systems": { "title": "Cerebras Systems", "type": "organization", "words": 822, "refs": 6, "outbound": [ "Cerebras-GPT", "Chinchilla", "Distributed_training", "EleutherAI", "GPT-3", "Hugging_Face", "Large_language_model", "Llama_3.1", "Meta_AI", "Mistral_AI", "NVIDIA", "Pretraining", "Scaling_laws", "Stated_missions_of_AI_labs", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Organizations", "Hardware and compute" ] }, "Chain-of-thought_prompting": { "title": "Chain-of-thought prompting", "type": "technique", "words": 563, "refs": 5, "outbound": [ "Anthropic", "Claude_(model_family)", "DeepSeek", "GPT-3", "GPT_(model_family)", "Google_DeepMind", "Large_language_model", "MMLU", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute" ], "categories": [ "Concepts", "Inference", "Reasoning models" ] }, "ChatGLM": { "title": "ChatGLM", "type": "model", "words": 398, "refs": 3, "outbound": [ "Alpaca", "Baichuan", "DeepSeek_(model_family)", "GLM-130B", "GLM-4", "GLM-4.5", "GLM_(model_family)", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "Pretraining", "Qwen_1", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "Vicuna", "Zhipu_AI" ], "categories": [ "Zhipu AI models", "Chinese AI labs", "Models", "Open-weight models", "2023 model releases" ] }, "Chinchilla": { "title": "Chinchilla", "type": "model", "words": 390, "refs": 4, "outbound": [ "Anthropic", "DeepSeek", "GPT-3", "Gemini_(model_family)", "Google_DeepMind", "Gopher", "LLaMA", "MMLU", "Megatron-Turing_NLG", "Meta_AI", "PaLM_2", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2022 model releases" ] }, "Chinese_AI_labs": { "title": "Chinese AI labs", "type": "reference", "words": 906, "refs": 3, "outbound": [ "01.AI", "Ant_Group", "Baichuan", "Baidu", "ByteDance", "Claude_Opus_4.5", "DeepSeek", "DeepSeek-R1", "DeepSeek_(model_family)", "DeepSeek_R1_release_shock", "Doubao", "ERNIE_Bot", "GLM_(model_family)", "GPT-4", "GPT-5", "Hailuo", "Huawei", "Hugging_Face", "Hunyuan", "Hunyuan_3.0", "InternLM", "InternVL", "Kimi_K2", "Kimi_K3", "LLaMA", "Ling_(models)", "Llama_(model_family)", "LongCat", "Meituan", "Meta_AI", "MiMo", "MiniMax", "MiniMax-M1", "MiniMax-M2", "Moonshot_AI", "NVIDIA", "OpenAI", "Pretraining", "Qwen3", "Qwen_(model_family)", "Qwen_1", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Seed-OSS", "Seedream", "SenseNova", "SenseTime", "Shanghai_AI_Laboratory", "Stated_missions_of_AI_labs", "StepFun", "Tencent", "Xiaomi", "Yi_(model_family)", "Zhipu_AI", "iFlytek", "iFlytek_Spark" ], "categories": [ "Chinese AI labs", "Organizations" ] }, "Circuits_(interpretability)": { "title": "Circuits (interpretability)", "type": "concept", "words": 2737, "refs": 20, "outbound": [ "Activation_steering", "Activation_verbalizer", "Anthropic", "Attention_sink", "CLIP", "Chain-of-thought_prompting", "Claude_(model_family)", "Claude_3", "Claude_3.5_Haiku", "Claude_Opus_4.6", "Compositional_generalization", "Constitutional_AI", "EleutherAI", "Eliciting_Latent_Knowledge", "Feed-forward_network", "GPT-2", "GPT-4", "Gemma", "Google_DeepMind", "In-context_learning", "Large_language_model", "Logit_lens", "Mechanistic_interpretability", "Multi-head_attention", "OpenAI", "Pythia", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Self-attention", "Softmax", "Sparse_autoencoder", "Superposition_(interpretability)", "Test-time_compute", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Concepts", "Alignment and safety" ] }, "Claude_(model_family)": { "title": "Claude (model family)", "type": "family", "words": 467, "refs": 3, "outbound": [ "Anthropic", "Claude_1", "Claude_2", "Claude_3", "Claude_3.5_Haiku", "Claude_3.5_Sonnet", "Claude_3.7_Sonnet", "Claude_4", "Claude_Fable_5", "Claude_Haiku_4.5", "Claude_Instant", "Claude_Opus_4.1", "Claude_Opus_4.5", "Claude_Opus_4.6", "Claude_Opus_5", "Claude_Sonnet_4.5", "Claude_Sonnet_5", "Constitutional_AI", "GPT-4", "GPT_(model_family)", "Large_language_model", "MMLU", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Model families", "Anthropic models" ] }, "Claude_1": { "title": "Claude 1", "type": "model", "words": 358, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_2", "Claude_Instant", "Constitutional_AI", "GPT-3.5", "GPT-4", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "2023 model releases" ] }, "Claude_2": { "title": "Claude 2", "type": "model", "words": 378, "refs": 4, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_1", "Claude_3", "Claude_Instant", "Constitutional_AI", "GPT-3.5", "GPT-4", "Gemini_(model_family)", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "2023 model releases" ] }, "Claude_3.5_Haiku": { "title": "Claude 3.5 Haiku", "type": "model", "words": 394, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3", "Claude_3.5_Sonnet", "Claude_3.7_Sonnet", "Claude_4", "Claude_Haiku_4.5", "Constitutional_AI", "GPT-4o", "GPT-4o_mini", "Gemini_1.5", "Large_language_model", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "2024 model releases" ] }, "Claude_3.5_Sonnet": { "title": "Claude 3.5 Sonnet", "type": "model", "words": 294, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3", "Claude_3.5_Haiku", "Claude_3.7_Sonnet", "Constitutional_AI", "GPT-4o", "Gemini_(model_family)", "MMLU", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "2024 model releases" ] }, "Claude_3.7_Sonnet": { "title": "Claude 3.7 Sonnet", "type": "model", "words": 286, "refs": 3, "outbound": [ "Anthropic", "Chain-of-thought_prompting", "Claude_(model_family)", "Claude_3.5_Sonnet", "Claude_4", "Constitutional_AI", "DeepSeek_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_3": { "title": "Claude 3", "type": "model", "words": 293, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_2", "Claude_3.5_Sonnet", "Constitutional_AI", "GPT-4", "GPT-4o", "Gemini_(model_family)", "Google_DeepMind", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Multimodal models", "2024 model releases" ] }, "Claude_4": { "title": "Claude 4", "type": "model", "words": 306, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3.5_Sonnet", "Claude_3.7_Sonnet", "Claude_Opus_4.1", "Claude_Opus_4.5", "Constitutional_AI", "Gemini_2.5", "Google_DeepMind", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_Fable_5": { "title": "Claude Fable 5", "type": "model", "words": 304, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_4", "Claude_Opus_4.8", "Claude_Sonnet_5", "Constitutional_AI", "GPT-5.6", "Gemini_3.5", "Kimi_K3", "Moonshot_AI", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2026 model releases" ] }, "Claude_Haiku_4.5": { "title": "Claude Haiku 4.5", "type": "model", "words": 392, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3", "Claude_3.5_Haiku", "Claude_4", "Claude_Instant", "Claude_Opus_4.1", "Claude_Opus_4.5", "Claude_Sonnet_4.5", "Constitutional_AI", "GPT-5", "Google_DeepMind", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_Instant": { "title": "Claude Instant", "type": "model", "words": 343, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_1", "Claude_2", "Claude_3", "Claude_3.5_Haiku", "Claude_Haiku_4.5", "Constitutional_AI", "GPT-3.5", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "2023 model releases" ] }, "Claude_Opus_4.1": { "title": "Claude Opus 4.1", "type": "model", "words": 370, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_4", "Claude_Opus_4.5", "Claude_Opus_4.6", "Claude_Sonnet_4.5", "Constitutional_AI", "GPT-5", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_Opus_4.5": { "title": "Claude Opus 4.5", "type": "model", "words": 294, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_4", "Claude_Opus_4.6", "Constitutional_AI", "DeepSeek_(model_family)", "GPT-5.1", "Gemini_3", "Google_DeepMind", "Kimi_(model_family)", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_Opus_4.6": { "title": "Claude Opus 4.6", "type": "model", "words": 379, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_4", "Claude_Haiku_4.5", "Claude_Opus_4.1", "Claude_Opus_4.5", "Claude_Sonnet_4.5", "Claude_Sonnet_5", "Constitutional_AI", "GPT-5.1", "GPT-5.2", "Gemini_3", "Kimi_K3", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2026 model releases" ] }, "Claude_Opus_4.8": { "title": "Claude Opus 4.8", "type": "model", "words": 399, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_Fable_5", "Claude_Opus_4.5", "Claude_Opus_4.6", "Claude_Opus_5", "Claude_Sonnet_5", "Constitutional_AI", "GPT-5.2", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2026 model releases" ] }, "Claude_Opus_5": { "title": "Claude Opus 5", "type": "model", "words": 698, "refs": 6, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_Fable_5", "Claude_Opus_4.8", "Claude_Sonnet_5", "Constitutional_AI", "Large_language_model", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2026 model releases" ] }, "Claude_Sonnet_4.5": { "title": "Claude Sonnet 4.5", "type": "model", "words": 396, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3.7_Sonnet", "Claude_4", "Claude_Haiku_4.5", "Claude_Opus_4.1", "Claude_Opus_4.5", "Claude_Sonnet_5", "Constitutional_AI", "GPT-5", "Gemini_2.5", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2025 model releases" ] }, "Claude_Sonnet_5": { "title": "Claude Sonnet 5", "type": "model", "words": 358, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Claude_3.5_Sonnet", "Claude_Haiku_4.5", "Claude_Opus_4.5", "Claude_Opus_4.6", "Claude_Opus_4.8", "Claude_Sonnet_4.5", "Constitutional_AI", "GPT-5", "Gemini_3", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Anthropic models", "Models", "Reasoning models", "2026 model releases" ] }, "Code_Llama": { "title": "Code Llama", "type": "model", "words": 360, "refs": 3, "outbound": [ "Alpaca", "Codestral", "DeepSeek-Coder-V2", "GPT-4", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "Llama_(model_family)", "Llama_2", "Llama_3.1", "Meta_AI", "Pretraining", "Qwen3-Coder", "Transformer_(architecture)", "Vicuna" ], "categories": [ "Meta models", "Models", "Open-weight models", "2023 model releases" ] }, "Codestral": { "title": "Codestral", "type": "model", "words": 375, "refs": 3, "outbound": [ "Code_Llama", "DeepSeek-Coder-V2", "Devstral", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Magistral", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mixtral_8x7B", "Qwen3-Coder", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "2024 model releases" ] }, "Codex_(2021_model)": { "title": "Codex (2021 model)", "type": "model", "words": 280, "refs": 3, "outbound": [ "Code_Llama", "Codestral", "Cursor", "DeepSeek-Coder-V2", "GPT-3", "GPT-4", "GPT_(model_family)", "OpenAI", "Qwen3-Coder", "StarCoder", "Transformer_(architecture)" ], "categories": [ "Models", "OpenAI models", "2021 model releases" ] }, "Cohere": { "title": "Cohere", "type": "organization", "words": 1374, "refs": 10, "outbound": [ "AI21_Labs", "Aleph_Alpha", "Anthropic", "Aya", "Chinese_AI_labs", "Command_(models)", "GPT-4o", "Google_DeepMind", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Meta_AI", "Mistral_AI", "Mixture_of_experts", "NVIDIA", "OpenAI", "Stated_missions_of_AI_labs", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs", "Open-source AI" ] }, "Comma_(models)": { "title": "Comma (models)", "type": "model", "words": 376, "refs": 4, "outbound": [ "Ai2", "Common_Pile", "DeepSeek_LLM", "EleutherAI", "GPT-NeoX-20B", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "Llama_2", "Llama_3", "MMLU", "NVIDIA", "Pretraining", "Pythia", "The_Pile", "Transformer_(architecture)" ], "categories": [ "2025 model releases", "Open-weight models", "Models" ] }, "Command_(models)": { "title": "Command (models)", "type": "model", "words": 329, "refs": 3, "outbound": [ "Aya", "Cohere", "EleutherAI", "GPT-4o", "Hugging_Face", "Llama_(model_family)", "MMLU", "Meta_AI", "Mistral_AI", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2024 model releases" ] }, "Common_Crawl": { "title": "Common Crawl", "type": "dataset", "words": 313, "refs": 4, "outbound": [ "Common_Pile", "EleutherAI", "Falcon_(models)", "GPT-3", "Large_language_model", "OLMo", "Pretraining", "T5", "The_Pile" ], "categories": [ "Training datasets", "Open-source AI" ] }, "Common_Pile": { "title": "Common Pile", "type": "dataset", "words": 564, "refs": 4, "outbound": [ "Ai2", "Comma_(models)", "EleutherAI", "GPT-J", "GPT-Neo", "GPT-NeoX-20B", "Hugging_Face", "LLaMA", "Large_language_model", "Llama_2", "Pythia", "The_Pile" ], "categories": [ "Training datasets", "Open-source AI" ] }, "Compositional_generalization": { "title": "Compositional generalization", "type": "concept", "words": 472, "refs": 7, "outbound": [ "Analytic_distillation", "In-context_learning", "Large_language_model", "Mechanistic_interpretability", "Ndea", "Next-token_prediction", "Scaling_laws", "Seq2seq", "Test-time_compute" ], "categories": [ "Concepts", "Training methods" ] }, "Constitutional_AI": { "title": "Constitutional AI", "type": "technique", "words": 349, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Meta_AI", "OpenAI", "Reinforcement_learning_from_human_feedback" ], "categories": [ "Post-training", "Alignment and safety" ] }, "Cross-entropy_loss": { "title": "Cross-entropy loss", "type": "concept", "words": 1249, "refs": 7, "outbound": [ "BERT", "Backpropagation", "Chinchilla", "Direct_preference_optimization", "Knowledge_distillation", "Large_language_model", "Masked_language_modeling", "Next-token_prediction", "OPT", "Pretraining", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Seq2seq", "Softmax", "Teacher_forcing", "Tokenization" ], "categories": [ "Concepts", "Training methods" ] }, "Cursor": { "title": "Cursor", "type": "organization", "words": 994, "refs": 8, "outbound": [ "Anthropic", "Claude_(model_family)", "GPT_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "Grok_(model_family)", "Large_language_model", "Mixture_of_experts", "NVIDIA", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Stated_missions_of_AI_labs", "xAI" ], "categories": [ "Organizations", "Inference" ] }, "DALL-E": { "title": "DALL-E", "type": "model", "words": 355, "refs": 3, "outbound": [ "CLIP", "DALL-E_2", "DALL-E_3", "GPT-2", "GPT-3", "GPT-4o", "Imagen", "Large_language_model", "OpenAI", "Sora", "Stability_AI", "Stable_Diffusion", "Transformer_(architecture)", "Veo", "Whisper" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2021 model releases" ] }, "DALL-E_2": { "title": "DALL-E 2", "type": "model", "words": 378, "refs": 3, "outbound": [ "CLIP", "DALL-E", "DALL-E_3", "FLUX", "GPT-3", "GPT-4o", "Imagen", "OpenAI", "Sora", "Stability_AI", "Stable_Diffusion", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2022 model releases" ] }, "DALL-E_3": { "title": "DALL-E 3", "type": "model", "words": 367, "refs": 4, "outbound": [ "CLIP", "DALL-E", "DALL-E_2", "FLUX", "GPT-4_Turbo", "GPT-4o", "Imagen", "OpenAI", "Sora", "Stable_Diffusion", "T5", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2023 model releases" ] }, "DBRX": { "title": "DBRX", "type": "model", "words": 403, "refs": 3, "outbound": [ "Databricks", "DeepSeek-V2", "Grok-1", "Hugging_Face", "Instruction_tuning", "Jamba", "Large_language_model", "Llama_2", "Llama_3", "Mixtral_8x22B", "Mixtral_8x7B", "Mixture_of_experts", "MosaicML", "NVIDIA", "Transformer_(architecture)" ], "categories": [ "Models", "Mixture-of-experts models", "Open-weight models", "2024 model releases" ] }, "Databricks": { "title": "Databricks", "type": "organization", "words": 820, "refs": 6, "outbound": [ "DBRX", "Grok-1", "Hugging_Face", "Large_language_model", "Llama_2", "MMLU", "MPT_(models)", "Meta_AI", "Mistral_AI", "Mixtral_8x7B", "Mixture_of_experts", "MosaicML", "NVIDIA", "OpenAI", "Stated_missions_of_AI_labs", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "David_Rumelhart": { "title": "David Rumelhart", "type": "person", "words": 1352, "refs": 13, "outbound": [ "AlexNet", "Backpropagation", "Boltzmann_machine", "Compositional_generalization", "Feed-forward_network", "Frank_Rosenblatt", "Geoffrey_Hinton", "Hopfield_network", "ImageNet", "John_Hopfield", "Large_language_model", "NETtalk", "Paul_Werbos", "Perceptron", "Walter_Pitts", "Warren_McCulloch" ], "categories": [ "People", "Computer science pioneers" ] }, "David_Silver": { "title": "David Silver", "type": "person", "words": 2020, "refs": 25, "outbound": [ "AlexNet", "AlphaFold", "AlphaGo", "AlphaZero", "Arthur_Samuel", "Backpropagation", "David_Rumelhart", "Demis_Hassabis", "Gemini_(model_family)", "Geoffrey_Hinton", "Google_DeepMind", "Ilya_Sutskever", "ImageNet", "John_Jumper", "Large_language_model", "OpenAI", "Paul_Werbos", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Richard_Sutton", "Safe_Superintelligence", "TD-Gammon", "Yann_LeCun", "o1" ], "categories": [ "People", "Machine learning researchers" ] }, "DeepSeek-Coder-V2": { "title": "DeepSeek-Coder-V2", "type": "model", "words": 351, "refs": 3, "outbound": [ "Code_Llama", "Codestral", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek_(model_family)", "DeepSeek_LLM", "GPT-4", "Hugging_Face", "Instruction_tuning", "Mixture_of_experts", "Multi-head_latent_attention", "Qwen2", "Qwen3-Coder", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "DeepSeek models", "Models", "2024 model releases", "Mixture-of-experts models", "Open-weight models" ] }, "DeepSeek-R1": { "title": "DeepSeek-R1", "type": "model", "words": 337, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "DeepSeek", "DeepSeek-V3", "DeepSeek_(model_family)", "DeepSeek_R1_release_shock", "Knowledge_distillation", "Llama_(model_family)", "Mixture_of_experts", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Scaling_laws", "Supervised_fine-tuning", "Test-time_compute", "o1" ], "categories": [ "DeepSeek models", "Models", "Reasoning models", "Open-weight models", "2025 model releases" ] }, "DeepSeek-V2": { "title": "DeepSeek-V2", "type": "model", "words": 382, "refs": 4, "outbound": [ "Baidu", "Chinese_AI_labs", "DeepSeek", "DeepSeek-Coder-V2", "DeepSeek-R1", "DeepSeek-V3", "DeepSeek_(model_family)", "DeepSeek_LLM", "GPT-4", "MMLU", "Mistral_AI", "Mixture_of_experts", "Multi-head_attention", "Multi-head_latent_attention", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "DeepSeek models", "Models", "Open-weight models", "Mixture-of-experts models", "2024 model releases" ] }, "DeepSeek-V3.1": { "title": "DeepSeek-V3.1", "type": "model", "words": 367, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek_(model_family)", "Hugging_Face", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Multi-head_latent_attention", "Qwen3", "Sparse_attention", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "DeepSeek models", "Models", "2025 model releases", "Mixture-of-experts models", "Open-weight models", "Reasoning models" ] }, "DeepSeek-V3": { "title": "DeepSeek-V3", "type": "model", "words": 306, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3.1", "DeepSeek_(model_family)", "DeepSeek_R1_release_shock", "Distributed_training", "GPT-4o", "MMLU", "Mixture_of_experts", "Multi-head_latent_attention", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "DeepSeek models", "Models", "Open-weight models", "Mixture-of-experts models", "2024 model releases" ] }, "DeepSeek": { "title": "DeepSeek", "type": "organization", "words": 599, "refs": 6, "outbound": [ "Anthropic", "Chinese_AI_labs", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek_(model_family)", "DeepSeek_LLM", "DeepSeek_R1_release_shock", "Google_DeepMind", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "MMLU", "Meta_AI", "Mistral_AI", "Mixture_of_experts", "Multi-head_latent_attention", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "o1" ], "categories": [ "Organizations", "Frontier labs", "Chinese AI labs" ] }, "DeepSeek_(model_family)": { "title": "DeepSeek (model family)", "type": "family", "words": 310, "refs": 3, "outbound": [ "DeepSeek", "DeepSeek-Coder-V2", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek_LLM", "GPT-4", "Llama_(model_family)", "MMLU", "Mistral_AI", "Mixture_of_experts", "Multi-head_latent_attention", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Sparse_attention", "Transformer_(architecture)", "o1" ], "categories": [ "Model families", "DeepSeek models", "Open-weight models" ] }, "DeepSeek_LLM": { "title": "DeepSeek LLM", "type": "model", "words": 348, "refs": 3, "outbound": [ "Chinchilla", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek_(model_family)", "Direct_preference_optimization", "GPT-3.5", "Instruction_tuning", "LLaMA", "Llama_2", "MMLU", "Mixture_of_experts", "Qwen_1", "RMSNorm", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "DeepSeek models", "Models", "2023 model releases", "Open-weight models" ] }, "DeepSeek_R1_release_shock": { "title": "DeepSeek R1 release shock", "type": "incident", "words": 363, "refs": 3, "outbound": [ "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "DeepSeek_(model_family)", "GLM_(model_family)", "Kimi_(model_family)", "Knowledge_distillation", "NVIDIA", "OpenAI", "Qwen_(model_family)", "Test-time_compute", "o1" ], "categories": [ "Incidents and controversies", "AI governance", "2025 model releases" ] }, "Demis_Hassabis": { "title": "Demis Hassabis", "type": "person", "words": 2258, "refs": 32, "outbound": [ "AlexNet", "AlphaFold", "AlphaGo", "AlphaZero", "Arthur_Samuel", "Backpropagation", "Chinchilla", "David_Silver", "Gemini_(model_family)", "Gemini_3", "Gemma", "Genie", "Geoffrey_Hinton", "Google_DeepMind", "ImageNet", "John_Hopfield", "John_Jumper", "Large_language_model", "PaLM", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Richard_Sutton", "Scaling_laws", "TD-Gammon", "Transformer_(architecture)", "o1" ], "categories": [ "People", "Machine learning researchers" ] }, "Devstral": { "title": "Devstral", "type": "model", "words": 366, "refs": 3, "outbound": [ "Code_Llama", "Codestral", "DeepSeek-Coder-V2", "Hugging_Face", "Magistral", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Medium_3", "Mistral_Small", "Mixtral_8x7B", "Pixtral", "Qwen3-Coder", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "Open-weight models", "2025 model releases" ] }, "Diffusion_language_model": { "title": "Diffusion language model", "type": "technique", "words": 1808, "refs": 8, "outbound": [ "BERT", "Chain-of-thought_prompting", "DALL-E_2", "FlashAttention", "GPT-2", "Gemini_(model_family)", "Gemini_2.0", "Google_DeepMind", "Hybrid_architecture_(LLM)", "Imagen", "In-context_learning", "Large_language_model", "Llama_3", "Masked_language_modeling", "NVIDIA", "Pretraining", "RoBERTa", "Scaling_laws", "Self-attention", "Stable_Diffusion", "State-space_model", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Architectures", "Concepts", "Inference" ] }, "Direct_preference_optimization": { "title": "Direct preference optimization", "type": "technique", "words": 342, "refs": 3, "outbound": [ "Constitutional_AI", "Hugging_Face", "Large_language_model", "Llama_3", "Meta_AI", "Mistral_AI", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Reward_model", "Supervised_fine-tuning" ], "categories": [ "Post-training", "Training methods" ] }, "DistilBERT": { "title": "DistilBERT", "type": "model", "words": 369, "refs": 3, "outbound": [ "ALBERT", "BERT", "ELECTRA", "GPT-1", "GPT-2", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "Masked_language_modeling", "RoBERTa", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Hugging Face models", "Open-weight models", "2019 model releases", "Models" ] }, "Distributed_training": { "title": "Distributed training", "type": "concept", "words": 313, "refs": 3, "outbound": [ "Anthropic", "DeepSeek", "GPT-3", "GPT-4", "Google_DeepMind", "Large_language_model", "Megatron-LM", "Meta_AI", "Mixture_of_experts", "NVIDIA", "OpenAI", "Transformer_(architecture)" ], "categories": [ "Training methods", "Hardware and compute" ] }, "Doubao": { "title": "Doubao", "type": "model", "words": 418, "refs": 4, "outbound": [ "Baidu", "ByteDance", "Chinese_AI_labs", "DeepSeek", "MMLU", "Mixture_of_experts", "Moonshot_AI", "NVIDIA", "Qwen_team", "Reinforcement_learning_from_human_feedback", "Seed-OSS", "Seedream", "Supervised_fine-tuning", "Tencent", "Test-time_compute", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Chinese AI labs", "Models", "2024 model releases" ] }, "Dropout": { "title": "Dropout", "type": "concept", "words": 1271, "refs": 6, "outbound": [ "Adam_(optimizer)", "BERT", "Embedding_(machine_learning)", "GPT-2", "ImageNet", "Layer_normalization", "Pretraining", "Scaling_laws", "Softmax", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Concepts", "Training methods" ] }, "ELECTRA": { "title": "ELECTRA", "type": "model", "words": 424, "refs": 3, "outbound": [ "ALBERT", "BERT", "DistilBERT", "GPT-1", "GPT-2", "GPT-3", "Hugging_Face", "Large_language_model", "Masked_language_modeling", "RoBERTa", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2020 model releases", "Open-weight models" ] }, "ELMo": { "title": "ELMo", "type": "model", "words": 443, "refs": 4, "outbound": [ "Ai2", "BERT", "GPT-1", "GloVe", "LSTM", "Large_language_model", "OLMo", "Pretraining", "RoBERTa", "Softmax", "Transformer_(architecture)", "Word2vec" ], "categories": [ "2018 model releases", "Models", "Open-weight models" ] }, "ERNIE_5": { "title": "ERNIE 5", "type": "model", "words": 402, "refs": 4, "outbound": [ "Baidu", "DeepSeek-V2", "ERNIE_Bot", "GLM-5", "GPT-5", "Gemini_2.5", "Hunyuan_3.0", "Kimi_K3", "Large_language_model", "Mixture_of_experts", "Qwen3-Max", "Transformer_(architecture)" ], "categories": [ "Baidu models", "Models", "Multimodal models", "Mixture-of-experts models", "2025 model releases", "2026 model releases" ] }, "ERNIE_Bot": { "title": "ERNIE Bot", "type": "model", "words": 389, "refs": 4, "outbound": [ "Baidu", "ChatGLM", "Chinese_AI_labs", "DeepSeek-R1", "DeepSeek-V2", "Doubao", "ERNIE_5", "GPT-4", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Qwen2.5", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "iFlytek_Spark", "o1" ], "categories": [ "Baidu models", "Chinese AI labs", "Models", "2023 model releases" ] }, "EleutherAI": { "title": "EleutherAI", "type": "organization", "words": 522, "refs": 4, "outbound": [ "Ai2", "Anthropic", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Hugging_Face", "Large_language_model", "MMLU", "Mechanistic_interpretability", "Meta_AI", "OLMo", "OpenAI", "Pretraining", "Pythia", "Stability_AI", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Eliciting_Latent_Knowledge": { "title": "Eliciting Latent Knowledge", "type": "concept", "words": 1440, "refs": 8, "outbound": [ "Activation_steering", "Activation_verbalizer", "Anthropic", "Chain-of-thought_prompting", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_Opus_4.6", "Constitutional_AI", "GPT-4", "Google_DeepMind", "Large_language_model", "Logit_lens", "Mechanistic_interpretability", "OpenAI", "Reinforcement_learning_from_human_feedback", "Sparse_autoencoder", "Superposition_(interpretability)" ], "categories": [ "Alignment and safety", "Concepts" ] }, "Embedding_(machine_learning)": { "title": "Embedding (machine learning)", "type": "concept", "words": 1605, "refs": 14, "outbound": [ "ALBERT", "BERT", "Cross-entropy_loss", "ELMo", "Feed-forward_network", "GPT-2", "GPT-3", "GPT_(model_family)", "Gemma", "GloVe", "Large_language_model", "Layer_normalization", "Llama_(model_family)", "Llama_3", "Logit_lens", "Next-token_prediction", "Positional_encoding", "Pretraining", "Self-attention", "Softmax", "Tokenization", "Transformer_(architecture)", "Word2vec" ], "categories": [ "Concepts", "Architectures" ] }, "FLUX": { "title": "FLUX", "type": "model", "words": 387, "refs": 3, "outbound": [ "CLIP", "DALL-E_3", "Grok-2", "Hugging_Face", "Imagen", "Kling", "Large_language_model", "Seedream", "Stability_AI", "Stable_Diffusion", "SynthID", "T5", "Transformer_(architecture)", "xAI" ], "categories": [ "Models", "2024 model releases", "Open-weight models", "Multimodal models" ] }, "Falcon_(models)": { "title": "Falcon (models)", "type": "model", "words": 337, "refs": 3, "outbound": [ "Chinese_AI_labs", "Hugging_Face", "Jamba", "Llama_(model_family)", "Llama_2", "MMLU", "Qwen_(model_family)", "Self-attention", "State-space_model", "Technology_Innovation_Institute", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2023 model releases" ] }, "Feed-forward_network": { "title": "Feed-forward network", "type": "concept", "words": 1343, "refs": 11, "outbound": [ "BERT", "Backpropagation", "DeepSeek-V3", "DeepSeek_(model_family)", "Dropout", "GPT-1", "GPT_(model_family)", "LLaMA", "Large_language_model", "Layer_normalization", "Llama_(model_family)", "Mistral_7B", "Mixtral_8x7B", "Mixture_of_experts", "Perceptron", "Qwen_(model_family)", "RMSNorm", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Architectures", "Concepts" ] }, "Flamingo": { "title": "Flamingo", "type": "model", "words": 368, "refs": 3, "outbound": [ "CLIP", "Chinchilla", "GPT-3", "Gemini_(model_family)", "Google_DeepMind", "Hugging_Face", "IDEFICS", "Large_language_model", "Moondream_(model_family)", "PaLM", "PaliGemma", "Transformer_(architecture)" ], "categories": [ "Google models", "Multimodal models", "Models", "2022 model releases" ] }, "Flan-T5": { "title": "Flan-T5", "type": "model", "words": 371, "refs": 3, "outbound": [ "Alpaca", "Chain-of-thought_prompting", "GPT-3", "Google_DeepMind", "Hugging_Face", "InstructGPT", "Instruction_tuning", "Knowledge_distillation", "LaMDA", "Large_language_model", "MMLU", "OpenAI", "PaLM", "T5", "Transformer_(architecture)", "Tulu_3" ], "categories": [ "Google models", "Models", "Open-weight models", "2022 model releases" ] }, "FlashAttention": { "title": "FlashAttention", "type": "technique", "words": 2678, "refs": 16, "outbound": [ "Attention_sink", "BERT", "DeepSeek", "DeepSeek-V2", "Diffusion_language_model", "Distributed_training", "GPT-2", "GPT-3", "GPT-4", "Gemma", "Hugging_Face", "Hybrid_architecture_(LLM)", "Large_language_model", "Llama_2", "Llama_3.1", "Megatron-LM", "Meta_AI", "Mistral_7B", "Mistral_AI", "Mixtral_8x7B", "Mixture_of_experts", "MosaicML", "Multi-head_attention", "Multi-head_latent_attention", "NVIDIA", "PagedAttention", "Qwen2.5", "Scaling_laws", "Self-attention", "Sliding-window_attention", "Softmax", "Sparse_attention", "Stable_Diffusion", "State-space_model", "Test-time_compute", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Architectures", "Training methods", "Inference" ] }, "Frank_Rosenblatt": { "title": "Frank Rosenblatt", "type": "person", "words": 1383, "refs": 14, "outbound": [ "Arthur_Samuel", "Backpropagation", "Boltzmann_machine", "David_Rumelhart", "Feed-forward_network", "Geoffrey_Hinton", "Hopfield_network", "John_Hopfield", "Large_language_model", "NETtalk", "Paul_Werbos", "Perceptron", "Transformer_(architecture)", "Walter_Pitts", "Warren_McCulloch" ], "categories": [ "People", "Computer science pioneers" ] }, "François_Chollet": { "title": "François Chollet", "type": "person", "words": 1983, "refs": 20, "outbound": [ "AlexNet", "Chain-of-thought_prompting", "Compositional_generalization", "DeepSeek-R1", "GPT-3", "Gemma", "Google_DeepMind", "Hugging_Face", "ImageNet", "In-context_learning", "Keras", "Large_language_model", "Llama_(model_family)", "MMLU", "Ndea", "Next-token_prediction", "OpenAI", "Pretraining", "Safe_Superintelligence", "Scaling_laws", "Stated_missions_of_AI_labs", "Test-time_compute", "Transformer_(architecture)", "Vision_Transformer", "o3" ], "categories": [ "People", "Machine learning researchers" ] }, "GLM-130B": { "title": "GLM-130B", "type": "model", "words": 409, "refs": 3, "outbound": [ "BLOOM", "ChatGLM", "Chinese_AI_labs", "Feed-forward_network", "GLM-4", "GLM_(model_family)", "GPT-3", "GPT-NeoX-20B", "Hugging_Face", "LLaMA", "Large_language_model", "Layer_normalization", "MMLU", "Megatron-LM", "NVIDIA", "OPT", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Zhipu AI models", "Models", "Open-weight models", "2022 model releases" ] }, "GLM-4.5": { "title": "GLM-4.5", "type": "model", "words": 366, "refs": 3, "outbound": [ "ChatGLM", "DeepSeek-V3", "GLM-130B", "GLM-4", "GLM-4.6", "GLM-5", "GLM_(model_family)", "Hugging_Face", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Qwen3", "Supervised_fine-tuning", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Models", "2025 model releases", "Mixture-of-experts models", "Open-weight models", "Reasoning models", "Chinese AI labs" ] }, "GLM-4.6": { "title": "GLM-4.6", "type": "model", "words": 379, "refs": 3, "outbound": [ "ChatGLM", "Claude_4", "DeepSeek-V3", "GLM-130B", "GLM-4", "GLM-4.5", "GLM-5", "GLM_(model_family)", "Hugging_Face", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "PagedAttention", "Qwen3", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Chinese AI labs", "Models", "Mixture-of-experts models", "Open-weight models", "Reasoning models", "2025 model releases" ] }, "GLM-4": { "title": "GLM-4", "type": "model", "words": 366, "refs": 3, "outbound": [ "ChatGLM", "DeepSeek-V2", "GLM-130B", "GLM-4.5", "GLM-4.6", "GLM_(model_family)", "GPT-4", "Hugging_Face", "Instruction_tuning", "Large_language_model", "MMLU", "Qwen2", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Zhipu AI models", "Models", "Chinese AI labs", "2024 model releases", "Open-weight models" ] }, "GLM-5": { "title": "GLM-5", "type": "model", "words": 392, "refs": 3, "outbound": [ "ChatGLM", "Claude_Opus_4.6", "DeepSeek", "GLM-130B", "GLM-4.5", "GLM-4.6", "GLM_(model_family)", "GPT-5.1", "Huawei", "Hugging_Face", "Kimi_K3", "Large_language_model", "MiniMax-M2", "Mixture_of_experts", "NVIDIA", "Qwen3-Max", "Sparse_attention", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Zhipu AI models", "Models", "2026 model releases", "Mixture-of-experts models", "Open-weight models", "Reasoning models" ] }, "GLM_(model_family)": { "title": "GLM (model family)", "type": "family", "words": 302, "refs": 3, "outbound": [ "BERT", "ChatGLM", "DeepSeek", "DeepSeek_(model_family)", "GLM-130B", "GLM-4", "GLM-4.5", "GLM-4.6", "GLM-5", "Hugging_Face", "Kimi_(model_family)", "MMLU", "Mixture_of_experts", "Next-token_prediction", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)", "Zhipu_AI" ], "categories": [ "Model families", "Chinese AI labs", "Open-weight models" ] }, "GPT-1": { "title": "GPT-1", "type": "model", "words": 400, "refs": 3, "outbound": [ "Anthropic", "BERT", "GPT-2", "GPT-3", "GPT-3.5", "GPT-4", "GPT_(model_family)", "Instruction_tuning", "Large_language_model", "OpenAI", "Pretraining", "Scaling_laws", "Self-attention", "Supervised_fine-tuning", "Tokenization", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "2018 model releases", "Open-weight models" ] }, "GPT-2": { "title": "GPT-2", "type": "model", "words": 314, "refs": 3, "outbound": [ "EleutherAI", "GPT-1", "GPT-3", "GPT_(model_family)", "Grok-1", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Meta_AI", "Next-token_prediction", "OpenAI", "Scaling_laws", "Transformer_(architecture)", "xAI" ], "categories": [ "OpenAI models", "Models", "2019 model releases" ] }, "GPT-3.5": { "title": "GPT-3.5", "type": "model", "words": 356, "refs": 4, "outbound": [ "Alpaca", "Chinese_AI_labs", "DeepSeek-V2", "GPT-3", "GPT-4", "GPT_(model_family)", "Gemini_(model_family)", "InstructGPT", "Knowledge_distillation", "LLaMA", "Large_language_model", "MMLU", "Meta_AI", "OpenAI", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)", "Vicuna" ], "categories": [ "OpenAI models", "Models", "2022 model releases" ] }, "GPT-3": { "title": "GPT-3", "type": "model", "words": 363, "refs": 3, "outbound": [ "Common_Crawl", "GPT-2", "GPT-3.5", "GPT-4", "GPT_(model_family)", "In-context_learning", "InstructGPT", "Large_language_model", "MMLU", "Next-token_prediction", "OpenAI", "Reinforcement_learning_from_human_feedback", "Self-attention", "Tokenization", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "2020 model releases" ] }, "GPT-4.5": { "title": "GPT-4.5", "type": "model", "words": 457, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "Claude_3.7_Sonnet", "GPT-4", "GPT-4_Turbo", "GPT-4o", "GPT-4o_mini", "GPT-5", "GPT_(model_family)", "Instruction_tuning", "Large_language_model", "Llama_4", "OpenAI", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)", "o1", "o3" ], "categories": [ "OpenAI models", "Models", "2025 model releases" ] }, "GPT-4": { "title": "GPT-4", "type": "model", "words": 354, "refs": 3, "outbound": [ "GPT-3", "GPT-3.5", "GPT-4_Turbo", "GPT-4o", "GPT_(model_family)", "Google_DeepMind", "Large_language_model", "MMLU", "Mixture_of_experts", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2023 model releases" ] }, "GPT-4_Turbo": { "title": "GPT-4 Turbo", "type": "model", "words": 374, "refs": 3, "outbound": [ "DALL-E", "GPT-4", "GPT-4.5", "GPT-4o", "GPT-4o_mini", "GPT_(model_family)", "InstructGPT", "Instruction_tuning", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2023 model releases" ] }, "GPT-4o": { "title": "GPT-4o", "type": "model", "words": 321, "refs": 3, "outbound": [ "Anthropic", "Claude_(model_family)", "GPT-4", "GPT-4_Turbo", "GPT-5", "GPT_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "Large_language_model", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Tokenization", "Transformer_(architecture)", "o1" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2024 model releases" ] }, "GPT-4o_mini": { "title": "GPT-4o mini", "type": "model", "words": 341, "refs": 3, "outbound": [ "Claude_3.5_Haiku", "GPT-3.5", "GPT-4_Turbo", "GPT-4o", "GPT-5", "GPT_(model_family)", "Gemini_1.5", "Instruction_tuning", "Knowledge_distillation", "Large_language_model", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "o1", "o4-mini" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2024 model releases" ] }, "GPT-5.1": { "title": "GPT-5.1", "type": "model", "words": 413, "refs": 3, "outbound": [ "Claude_Opus_4.5", "GPT-4", "GPT-4o", "GPT-5", "GPT_(model_family)", "Gemini_3", "Grok_4.1", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "gpt-oss", "o1", "o3" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2025 model releases" ] }, "GPT-5.2": { "title": "GPT-5.2", "type": "model", "words": 369, "refs": 4, "outbound": [ "Claude_Opus_4.6", "GPT-4o_mini", "GPT-5", "GPT-5.1", "GPT-5.6", "GPT_(model_family)", "Gemini_3", "Google_DeepMind", "Grok_4.5", "Large_language_model", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)", "o1", "o3", "o4-mini", "xAI" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2025 model releases" ] }, "GPT-5.6": { "title": "GPT-5.6", "type": "model", "words": 389, "refs": 3, "outbound": [ "Claude_Opus_4.8", "GPT-4o", "GPT-4o_mini", "GPT-5", "GPT-5.1", "GPT-5.2", "GPT_(model_family)", "OpenAI", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)", "o3", "o4-mini" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2026 model releases" ] }, "GPT-5": { "title": "GPT-5", "type": "model", "words": 315, "refs": 3, "outbound": [ "Anthropic", "Claude_4", "DeepSeek_(model_family)", "GPT-2", "GPT-4o", "GPT-5.1", "GPT_(model_family)", "Gemini_2.5", "Google_DeepMind", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Test-time_compute", "gpt-oss", "o1", "o3" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2025 model releases" ] }, "GPT-J": { "title": "GPT-J", "type": "model", "words": 415, "refs": 3, "outbound": [ "BLOOM", "EleutherAI", "GPT-2", "GPT-3", "GPT-Neo", "GPT-NeoX-20B", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "OPT", "OpenAI", "PaLM", "Pythia", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Open-source AI", "2021 model releases" ] }, "GPT-Neo": { "title": "GPT-Neo", "type": "model", "words": 398, "refs": 3, "outbound": [ "BLOOM", "Distributed_training", "EleutherAI", "GPT-2", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Hugging_Face", "Instruction_tuning", "Large_language_model", "OPT", "OpenAI", "Pythia", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Open-source AI", "2021 model releases" ] }, "GPT-NeoX-20B": { "title": "GPT-NeoX-20B", "type": "model", "words": 353, "refs": 3, "outbound": [ "BLOOM", "Distributed_training", "EleutherAI", "GPT-2", "GPT-3", "GPT-J", "GPT-Neo", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Megatron-LM", "NVIDIA", "OLMo", "OPT", "Pythia", "Reinforcement_learning_from_human_feedback", "The_Pile", "Transformer_(architecture)" ], "categories": [ "2022 model releases", "Open-weight models", "Open-source AI", "Models" ] }, "GPT_(model_family)": { "title": "GPT (model family)", "type": "family", "words": 321, "refs": 5, "outbound": [ "BERT", "GPT-1", "GPT-2", "GPT-3", "GPT-3.5", "GPT-4", "GPT-4.5", "GPT-4_Turbo", "GPT-4o", "GPT-5", "GPT-5.1", "GPT-5.6", "Gemini_(model_family)", "Google_DeepMind", "In-context_learning", "InstructGPT", "Large_language_model", "MMLU", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "gpt-oss", "o1", "o3", "o4-mini" ], "categories": [ "Model families", "OpenAI models" ] }, "Gemini_(model_family)": { "title": "Gemini (model family)", "type": "family", "words": 318, "refs": 3, "outbound": [ "Chinchilla", "Claude_(model_family)", "GPT-4", "GPT_(model_family)", "Gemini_1.0", "Gemini_1.5", "Gemini_2.0", "Gemini_2.5", "Gemini_3", "Gemini_3.5", "Gemma", "Genie", "Google_DeepMind", "Gopher", "Imagen", "LaMDA", "Large_language_model", "MMLU", "OpenAI", "PaLM", "PaLM_2", "Transformer_(architecture)", "Veo" ], "categories": [ "Model families", "Google models" ] }, "Gemini_1.0": { "title": "Gemini 1.0", "type": "model", "words": 401, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "GPT-4", "Gemini_(model_family)", "Gemini_1.5", "Gemini_2.0", "Gemini_3", "Google_DeepMind", "Instruction_tuning", "LaMDA", "Large_language_model", "MMLU", "Mixture_of_experts", "OpenAI", "PaLM", "PaLM_2", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "Multimodal models", "2023 model releases" ] }, "Gemini_1.5": { "title": "Gemini 1.5", "type": "model", "words": 312, "refs": 3, "outbound": [ "Anthropic", "Claude_3", "GPT-4", "GPT-4o", "Gemini_(model_family)", "Gemini_1.0", "Gemini_2.0", "Gemini_2.5", "Google_DeepMind", "In-context_learning", "MMLU", "Mixture_of_experts", "OpenAI", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "Multimodal models", "2024 model releases" ] }, "Gemini_2.0": { "title": "Gemini 2.0", "type": "model", "words": 371, "refs": 4, "outbound": [ "Chain-of-thought_prompting", "Claude_3.5_Sonnet", "GPT-4o", "Gemini_(model_family)", "Gemini_1.5", "Gemini_2.5", "Google_DeepMind", "MMLU", "Mixture_of_experts", "OpenAI", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Google models", "Models", "Multimodal models", "2024 model releases" ] }, "Gemini_2.5": { "title": "Gemini 2.5", "type": "model", "words": 298, "refs": 3, "outbound": [ "Anthropic", "Chain-of-thought_prompting", "Claude_4", "Distributed_training", "GPT-4", "Gemini_(model_family)", "Gemini_2.0", "Gemini_3", "Gemma", "Google_DeepMind", "MMLU", "Mixture_of_experts", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Google models", "Models", "Reasoning models", "2025 model releases" ] }, "Gemini_3.5": { "title": "Gemini 3.5", "type": "model", "words": 389, "refs": 3, "outbound": [ "Claude_Opus_4.6", "GPT-5.1", "Gemini_(model_family)", "Gemini_1.0", "Gemini_1.5", "Gemini_2.5", "Gemini_3", "Google_DeepMind", "Large_language_model", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "Multimodal models", "Reasoning models", "2026 model releases" ] }, "Gemini_3": { "title": "Gemini 3", "type": "model", "words": 308, "refs": 3, "outbound": [ "Claude_Opus_4.5", "GPT-5.1", "Gemini_(model_family)", "Gemini_1.0", "Gemini_2.5", "Gemini_3.5", "Google_DeepMind", "MMLU", "Mixture_of_experts", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "Reasoning models", "2025 model releases" ] }, "Gemma": { "title": "Gemma", "type": "model", "words": 326, "refs": 4, "outbound": [ "Gemini_(model_family)", "Google_DeepMind", "Hugging_Face", "Keras", "Knowledge_distillation", "Llama_(model_family)", "Llama_2", "MMLU", "Mistral_(model_family)", "Mistral_AI", "OpenAI", "PaliGemma", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "Google models", "Models", "Open-weight models", "2024 model releases" ] }, "Genie": { "title": "Genie", "type": "model", "words": 408, "refs": 3, "outbound": [ "Gemini_(model_family)", "Google_DeepMind", "Imagen", "Large_language_model", "Movie_Gen", "Scaling_laws", "Sora", "Transformer_(architecture)", "Veo" ], "categories": [ "Google models", "Models", "Multimodal models", "2024 model releases", "2025 model releases" ] }, "Geoffrey_Hinton": { "title": "Geoffrey Hinton", "type": "person", "words": 2831, "refs": 28, "outbound": [ "Adam_(optimizer)", "AlexNet", "Alex_Krizhevsky", "Andrew_Ng", "Backpropagation", "Boltzmann_machine", "David_Rumelhart", "DistilBERT", "Dropout", "Embedding_(machine_learning)", "Frank_Rosenblatt", "Google_DeepMind", "Hopfield_network", "Ilya_Sutskever", "ImageNet", "John_Hopfield", "Knowledge_distillation", "Large_language_model", "Mixture_of_experts", "NETtalk", "Paul_Werbos", "Perceptron", "Pretraining", "Scaling_laws", "Softmax", "Word2vec", "Yann_LeCun", "Yoshua_Bengio" ], "categories": [ "People", "Machine learning researchers" ] }, "GloVe": { "title": "GloVe", "type": "model", "words": 424, "refs": 3, "outbound": [ "BERT", "Common_Crawl", "ELMo", "Embedding_(machine_learning)", "GPT-1", "Hugging_Face", "LSTM", "Large_language_model", "Pretraining", "Self-attention", "Transformer_(architecture)", "Word2vec" ], "categories": [ "Models", "Open-source AI" ] }, "Google_DeepMind": { "title": "Google DeepMind", "type": "organization", "words": 536, "refs": 5, "outbound": [ "BERT", "Chinchilla", "Demis_Hassabis", "GPT-3", "GPT-4", "Gemini_(model_family)", "Gemma", "Large_language_model", "MMLU", "OpenAI", "Reinforcement_learning_from_human_feedback", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] }, "Gopher": { "title": "Gopher", "type": "model", "words": 390, "refs": 3, "outbound": [ "Chinchilla", "GPT-3", "Gemini_(model_family)", "Google_DeepMind", "LaMDA", "Large_language_model", "MMLU", "Megatron-Turing_NLG", "PaLM", "Positional_encoding", "RMSNorm", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2021 model releases" ] }, "Grok-1.5": { "title": "Grok-1.5", "type": "model", "words": 337, "refs": 3, "outbound": [ "Claude_3", "Distributed_training", "GPT-4", "Gemini_1.5", "Grok-1", "Grok-2", "Grok_(model_family)", "Grok_3", "Large_language_model", "MMLU", "Mistral_Large", "Mixture_of_experts", "Transformer_(architecture)", "xAI" ], "categories": [ "xAI models", "Models", "2024 model releases" ] }, "Grok-1": { "title": "Grok-1", "type": "model", "words": 368, "refs": 4, "outbound": [ "Chinese_AI_labs", "GPT-3.5", "GPT-4", "Grok-1.5", "Grok_(model_family)", "Hugging_Face", "Kimi_K2", "MMLU", "Mixtral_8x7B", "Mixture_of_experts", "OpenAI", "Transformer_(architecture)", "xAI" ], "categories": [ "xAI models", "Models", "Open-weight models", "Mixture-of-experts models", "2023 model releases" ] }, "Grok-2": { "title": "Grok-2", "type": "model", "words": 381, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "FLUX", "GPT-4o", "Grok-1", "Grok-1.5", "Grok_(model_family)", "Grok_3", "Grok_4", "Hugging_Face", "Instruction_tuning", "Large_language_model", "MMLU", "Transformer_(architecture)", "xAI" ], "categories": [ "xAI models", "Models", "2024 model releases" ] }, "Grok_(model_family)": { "title": "Grok (model family)", "type": "family", "words": 289, "refs": 3, "outbound": [ "Anthropic", "GPT_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "Grok-1", "Grok-1.5", "Grok-2", "Grok_3", "Grok_4", "Grok_4.1", "Grok_4.5", "MMLU", "Mixture_of_experts", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)", "xAI" ], "categories": [ "Model families", "xAI models" ] }, "Grok_3": { "title": "Grok 3", "type": "model", "words": 397, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "DeepSeek-R1", "GPT-4o", "Gemini_2.0", "Grok-1", "Grok-2", "Grok_(model_family)", "Grok_4", "Large_language_model", "OpenAI", "Test-time_compute", "Transformer_(architecture)", "o1", "o3", "xAI" ], "categories": [ "xAI models", "Models", "Reasoning models", "2025 model releases" ] }, "Grok_4.1": { "title": "Grok 4.1", "type": "model", "words": 379, "refs": 3, "outbound": [ "Claude_Opus_4.5", "GPT-5.1", "Gemini_3", "Grok_(model_family)", "Grok_3", "Grok_4", "Grok_4.5", "Large_language_model", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "xAI" ], "categories": [ "xAI models", "Models", "Reasoning models", "2025 model releases" ] }, "Grok_4.5": { "title": "Grok 4.5", "type": "model", "words": 414, "refs": 3, "outbound": [ "Anthropic", "Chain-of-thought_prompting", "Claude_(model_family)", "Claude_Opus_4.5", "Cursor", "GPT-5.1", "Gemini_3.5", "Grok-1", "Grok_(model_family)", "Grok_3", "Grok_4", "Grok_4.1", "Large_language_model", "NVIDIA", "OpenAI", "Transformer_(architecture)", "xAI" ], "categories": [ "xAI models", "Models", "Reasoning models", "2026 model releases" ] }, "Grok_4": { "title": "Grok 4", "type": "model", "words": 388, "refs": 3, "outbound": [ "GPT-5", "Gemini_2.5", "Grok-1", "Grok_(model_family)", "Grok_3", "Grok_4.1", "Grok_4.5", "Large_language_model", "MMLU", "NVIDIA", "Test-time_compute", "Transformer_(architecture)", "o3", "xAI" ], "categories": [ "xAI models", "Models", "Reasoning models", "2025 model releases" ] }, "Hailuo": { "title": "Hailuo", "type": "model", "words": 366, "refs": 3, "outbound": [ "ByteDance", "Chinese_AI_labs", "Doubao", "HunyuanVideo", "Kimi_(model_family)", "Kling", "Large_language_model", "MiniMax", "MiniMax-M1", "MiniMax-Text-01", "Moonshot_AI", "Movie_Gen", "Seedream", "Sora", "Transformer_(architecture)", "Veo" ], "categories": [ "Models", "Multimodal models", "Chinese AI labs", "2024 model releases" ] }, "Hopfield_network": { "title": "Hopfield network", "type": "concept", "words": 1236, "refs": 12, "outbound": [ "Backpropagation", "Boltzmann_machine", "Frank_Rosenblatt", "Geoffrey_Hinton", "John_Hopfield", "Perceptron", "Self-attention", "Softmax", "Transformer_(architecture)", "Walter_Pitts", "Warren_McCulloch" ], "categories": [ "Concepts", "Architectures" ] }, "Huawei": { "title": "Huawei", "type": "organization", "words": 565, "refs": 5, "outbound": [ "Baidu", "ByteDance", "Chinese_AI_labs", "Distributed_training", "LongCat", "Meituan", "NVIDIA", "Qwen_(model_family)", "Zhipu_AI", "iFlytek" ], "categories": [ "Organizations", "Chinese AI labs", "Hardware and compute" ] }, "Hugging_Face": { "title": "Hugging Face", "type": "organization", "words": 530, "refs": 5, "outbound": [ "Anthropic", "BLOOM", "DeepSeek", "Direct_preference_optimization", "EleutherAI", "IDEFICS", "Large_language_model", "MMLU", "Meta_AI", "Mistral_AI", "OpenAI", "SmolLM", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Hunyuan": { "title": "Hunyuan", "type": "model", "words": 363, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "GLM-4.5", "Google_DeepMind", "Hugging_Face", "HunyuanVideo", "Hunyuan_3.0", "Llama_3.1", "MMLU", "Mixture_of_experts", "Qwen3", "Reinforcement_learning_from_human_feedback", "Stability_AI", "Stable_Diffusion", "Supervised_fine-tuning", "Tencent", "Transformer_(architecture)", "Veo" ], "categories": [ "Chinese AI labs", "Models", "Open-weight models", "Mixture-of-experts models", "2024 model releases" ] }, "HunyuanVideo": { "title": "HunyuanVideo", "type": "model", "words": 343, "refs": 3, "outbound": [ "CLIP", "Chinese_AI_labs", "DALL-E", "Hailuo", "Hugging_Face", "Hunyuan", "Hunyuan_3.0", "Large_language_model", "MiniMax", "Movie_Gen", "OpenAI", "Sora", "Stable_Diffusion", "T5", "Tencent", "Transformer_(architecture)", "Veo" ], "categories": [ "Tencent models", "2024 model releases", "Multimodal models", "Open-weight models", "Models" ] }, "Hunyuan_3.0": { "title": "Hunyuan 3.0", "type": "model", "words": 394, "refs": 3, "outbound": [ "Chinese_AI_labs", "DeepSeek-V3", "GLM-5", "Hugging_Face", "Hunyuan", "HunyuanVideo", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Qwen3", "Tencent", "Transformer_(architecture)" ], "categories": [ "Chinese AI labs", "Models", "Mixture-of-experts models", "Open-weight models", "Reasoning models", "2026 model releases" ] }, "Hybrid_architecture_(LLM)": { "title": "Hybrid architecture (LLM)", "type": "concept", "words": 1698, "refs": 12, "outbound": [ "AI21_Labs", "Attention_sink", "Diffusion_language_model", "FlashAttention", "Gemma", "Google_DeepMind", "IBM_Granite", "Jamba", "Large_language_model", "Llama_4", "MiniMax-M1", "MiniMax-Text-01", "Mistral_7B", "Mixture_of_experts", "Moonshot_AI", "Multi-head_latent_attention", "NVIDIA", "Nemotron", "OpenAI", "PagedAttention", "Qwen3", "Qwen_team", "Self-attention", "Sliding-window_attention", "Sparse_attention", "State-space_model", "Tokenization", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "Architectures", "Concepts" ] }, "IBM_Granite": { "title": "IBM Granite", "type": "model", "words": 361, "refs": 3, "outbound": [ "Gemma", "Hugging_Face", "Hybrid_architecture_(LLM)", "Instruction_tuning", "Jamba", "Mistral_7B", "Mixture_of_experts", "NVIDIA", "Nemotron", "RWKV", "StarCoder", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2024 model releases", "2025 model releases" ] }, "IDEFICS": { "title": "IDEFICS", "type": "model", "words": 312, "refs": 3, "outbound": [ "BLOOM", "Flamingo", "Google_DeepMind", "Hugging_Face", "Llama_(model_family)", "Llama_3", "Mistral_7B", "Moondream_(model_family)", "OLMo", "PaliGemma", "Pythia", "Qwen2.5-VL" ], "categories": [ "Models", "Multimodal models", "Open-weight models", "2023 model releases" ] }, "Ian_Goodfellow": { "title": "Ian Goodfellow", "type": "person", "words": 1122, "refs": 11, "outbound": [ "Andrew_Ng", "Backpropagation", "Cross-entropy_loss", "Dropout", "Google_DeepMind", "Ilya_Sutskever", "ImageNet", "Imagen", "OpenAI", "Stable_Diffusion", "Yoshua_Bengio" ], "categories": [ "People", "Machine learning researchers" ] }, "Ilya_Sutskever": { "title": "Ilya Sutskever", "type": "person", "words": 2002, "refs": 17, "outbound": [ "AlexNet", "Alex_Krizhevsky", "AlphaGo", "Andrew_Ng", "CLIP", "DALL-E", "Dropout", "GPT-2", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "ImageNet", "In-context_learning", "InstructGPT", "LSTM", "Large_language_model", "Next-token_prediction", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)", "Safe_Superintelligence", "Scaling_laws", "Seq2seq", "Transformer_(architecture)", "Word2vec", "o1" ], "categories": [ "People", "Machine learning researchers" ] }, "ImageNet": { "title": "ImageNet", "type": "dataset", "words": 1097, "refs": 14, "outbound": [ "AlexNet", "Alex_Krizhevsky", "Andrej_Karpathy", "Backpropagation", "CLIP", "Dropout", "Ilya_Sutskever", "NVIDIA", "OpenAI", "Perceptron", "Pretraining", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Training datasets", "Pre-transformer systems" ] }, "Imagen": { "title": "Imagen", "type": "model", "words": 410, "refs": 4, "outbound": [ "DALL-E", "Gemini_(model_family)", "Gemini_2.0", "Google_DeepMind", "Seedream", "Sora", "Stable_Diffusion", "SynthID", "T5", "Transformer_(architecture)", "Veo" ], "categories": [ "Google models", "Models", "Multimodal models", "2022 model releases" ] }, "In-context_learning": { "title": "In-context learning", "type": "concept", "words": 340, "refs": 4, "outbound": [ "Chain-of-thought_prompting", "Circuits_(interpretability)", "GPT-2", "GPT-3", "Large_language_model", "Mechanistic_interpretability", "Pretraining", "Test-time_compute" ], "categories": [ "Concepts", "Inference" ] }, "InstructGPT": { "title": "InstructGPT", "type": "model", "words": 291, "refs": 3, "outbound": [ "Anthropic", "Constitutional_AI", "Direct_preference_optimization", "GPT-3", "GPT-3.5", "GPT-4", "GPT_(model_family)", "Hugging_Face", "Llama_(model_family)", "OpenAI", "Reinforcement_learning_from_human_feedback", "Reward_model", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "OpenAI models", "Models", "2022 model releases", "Alignment and safety" ] }, "Instruction_tuning": { "title": "Instruction tuning", "type": "technique", "words": 288, "refs": 3, "outbound": [ "Alpaca", "Anthropic", "Constitutional_AI", "EleutherAI", "Flan-T5", "Hugging_Face", "InstructGPT", "LLaMA", "Large_language_model", "Llama_(model_family)", "MMLU", "Meta_AI", "Mistral_AI", "Next-token_prediction", "OpenAI", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Post-training", "Training methods" ] }, "InternLM": { "title": "InternLM", "type": "model", "words": 335, "refs": 3, "outbound": [ "ChatGLM", "Chinese_AI_labs", "DeepSeek_LLM", "Hugging_Face", "Instruction_tuning", "InternVL", "Large_language_model", "Llama_3.1", "Qwen2.5", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "SenseTime", "Shanghai_AI_Laboratory", "Transformer_(architecture)" ], "categories": [ "Chinese AI labs", "Open-weight models", "Models", "2023 model releases" ] }, "InternVL": { "title": "InternVL", "type": "model", "words": 404, "refs": 4, "outbound": [ "Chinese_AI_labs", "GPT-4", "GPT-4o", "Hugging_Face", "Instruction_tuning", "InternLM", "Large_language_model", "Molmo", "PaliGemma", "Qwen2.5", "Qwen2.5-VL", "SenseTime", "Shanghai_AI_Laboratory", "Transformer_(architecture)" ], "categories": [ "Chinese AI labs", "Multimodal models", "Open-weight models", "Models", "2023 model releases" ] }, "Jamba": { "title": "Jamba", "type": "model", "words": 344, "refs": 3, "outbound": [ "AI21_Labs", "Hugging_Face", "Hybrid_architecture_(LLM)", "Jurassic_(models)", "MMLU", "MiniMax", "Mixtral_8x7B", "Mixture_of_experts", "NVIDIA", "Nemotron", "Self-attention", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Mixture-of-experts models", "2024 model releases" ] }, "John_Hopfield": { "title": "John Hopfield", "type": "person", "words": 1661, "refs": 19, "outbound": [ "AlphaFold", "Backpropagation", "Boltzmann_machine", "David_Rumelhart", "Demis_Hassabis", "Frank_Rosenblatt", "Geoffrey_Hinton", "Hopfield_network", "John_Jumper", "NETtalk", "Perceptron", "Transformer_(architecture)", "Walter_Pitts", "Warren_McCulloch" ], "categories": [ "People", "Computer science pioneers" ] }, "John_Jumper": { "title": "John Jumper", "type": "person", "words": 1397, "refs": 17, "outbound": [ "AlphaFold", "AlphaGo", "Anthropic", "Backpropagation", "David_Silver", "Demis_Hassabis", "Google_DeepMind", "John_Hopfield", "Transformer_(architecture)" ], "categories": [ "People", "Machine learning researchers" ] }, "Jurassic_(models)": { "title": "Jurassic (models)", "type": "model", "words": 346, "refs": 3, "outbound": [ "AI21_Labs", "Amazon_Titan", "GPT-3", "Instruction_tuning", "Jamba", "Large_language_model", "Megatron-Turing_NLG", "OpenAI", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Models", "2021 model releases", "2023 model releases" ] }, "Keras": { "title": "Keras", "type": "concept", "words": 928, "refs": 12, "outbound": [ "BERT", "Falcon_(models)", "GPT-2", "Gemma", "Google_DeepMind", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Mistral_(model_family)", "Ndea", "PaliGemma", "Qwen3", "Segment_Anything", "Stable_Diffusion", "Supervised_fine-tuning", "gpt-oss" ], "categories": [ "Deep learning frameworks", "Open-source AI" ] }, "Kimi_(model_family)": { "title": "Kimi (model family)", "type": "family", "words": 320, "refs": 3, "outbound": [ "Anthropic", "DeepSeek", "DeepSeek_(model_family)", "GLM_(model_family)", "Kimi_K2", "Kimi_K3", "Kimi_k1.5", "MMLU", "Mixture_of_experts", "Moonshot_AI", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)", "o1" ], "categories": [ "Model families", "Chinese AI labs", "Open-weight models" ] }, "Kimi_K2": { "title": "Kimi K2", "type": "model", "words": 392, "refs": 4, "outbound": [ "Anthropic", "Claude_4", "DeepSeek", "DeepSeek-V3", "DeepSeek_R1_release_shock", "GPT_(model_family)", "Hugging_Face", "Kimi_(model_family)", "Kimi_K3", "Kimi_k1.5", "Mixture_of_experts", "Moonshot_AI", "OpenAI", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Chinese AI labs", "Models", "Open-weight models", "Mixture-of-experts models", "2025 model releases" ] }, "Kimi_K3": { "title": "Kimi K3", "type": "model", "words": 346, "refs": 3, "outbound": [ "Anthropic", "Chinese_AI_labs", "Claude_Fable_5", "DeepSeek", "GPT-5.6", "Kimi_(model_family)", "Kimi_K2", "LongCat", "Meituan", "Mixture_of_experts", "Moonshot_AI", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Chinese AI labs", "Models", "Open-weight models", "Mixture-of-experts models", "Reasoning models", "2026 model releases" ] }, "Kimi_k1.5": { "title": "Kimi k1.5", "type": "model", "words": 390, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "Claude_3.5_Sonnet", "DeepSeek", "DeepSeek-R1", "DeepSeek_R1_release_shock", "Direct_preference_optimization", "GPT-4o", "Kimi_(model_family)", "Kimi_K2", "Kimi_K3", "Moonshot_AI", "OpenAI", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Moonshot AI models", "Models", "Reasoning models", "Multimodal models", "2025 model releases" ] }, "Kling": { "title": "Kling", "type": "model", "words": 378, "refs": 3, "outbound": [ "ByteDance", "Chinese_AI_labs", "Google_DeepMind", "Hailuo", "HunyuanVideo", "Meta_AI", "MiniMax", "Movie_Gen", "OpenAI", "Seedream", "Sora", "Sora_2", "Tencent", "Transformer_(architecture)", "Veo" ], "categories": [ "Chinese AI labs", "Multimodal models", "Models", "2024 model releases" ] }, "Knowledge_distillation": { "title": "Knowledge distillation", "type": "technique", "words": 311, "refs": 3, "outbound": [ "Analytic_distillation", "DeepSeek", "DeepSeek-R1", "GPT-4", "GPT_(model_family)", "Gemini_(model_family)", "Gemma", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Llama_3.3", "OpenAI", "Qwen_(model_family)", "Scaling_laws" ], "categories": [ "Training methods", "Post-training", "Concepts" ] }, "LFM_(models)": { "title": "LFM (models)", "type": "model", "words": 562, "refs": 5, "outbound": [ "Falcon_(models)", "Gemma", "Hugging_Face", "Hybrid_architecture_(LLM)", "Jamba", "Knowledge_distillation", "Large_language_model", "Liquid_AI", "Mixture_of_experts", "Phi_(model_family)", "Qwen3", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models" ] }, "LLaMA": { "title": "LLaMA", "type": "model", "words": 315, "refs": 3, "outbound": [ "Alpaca", "Chinchilla", "DeepSeek", "EleutherAI", "GPT-3", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Llama_2", "MMLU", "Meta_AI", "Mistral_AI", "Positional_encoding", "Qwen_(model_family)", "RMSNorm", "Scaling_laws", "Transformer_(architecture)", "Vicuna" ], "categories": [ "Meta models", "Models", "Open-weight models", "2023 model releases" ] }, "LLaVA": { "title": "LLaVA", "type": "model", "words": 322, "refs": 3, "outbound": [ "Alpaca", "CLIP", "GPT-4", "Hugging_Face", "Instruction_tuning", "InternVL", "Knowledge_distillation", "Large_language_model", "MMLU", "Microsoft_AI", "Moondream_(model_family)", "Qwen_(model_family)", "Vicuna", "Vision_Transformer" ], "categories": [ "Models", "Open-weight models", "Vision-language models", "2023 model releases" ] }, "LSTM": { "title": "LSTM", "type": "concept", "words": 2197, "refs": 18, "outbound": [ "Andrej_Karpathy", "BERT", "Backpropagation", "Distributed_training", "Dropout", "ELMo", "GPT-2", "Google_DeepMind", "Large_language_model", "RWKV", "Scaling_laws", "Self-attention", "Seq2seq", "State-space_model", "The_Unreasonable_Effectiveness_of_Recurrent_Neural_Networks", "Transformer_(architecture)" ], "categories": [ "Architectures", "Concepts" ] }, "LaMDA": { "title": "LaMDA", "type": "model", "words": 397, "refs": 4, "outbound": [ "GPT-3", "GPT-3.5", "Gemini_(model_family)", "Gemini_1.0", "Google_DeepMind", "Instruction_tuning", "Large_language_model", "Meena", "PaLM", "PaLM_2", "Pretraining", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2021 model releases" ] }, "Large_language_model": { "title": "Large language model", "type": "concept", "words": 1972, "refs": 32, "outbound": [ "Adam_(optimizer)", "Anthropic", "BERT", "Backpropagation", "CLIP", "Chain-of-thought_prompting", "Chinchilla", "Cross-entropy_loss", "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "Diffusion_language_model", "Direct_preference_optimization", "ELMo", "Embedding_(machine_learning)", "GPT-2", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Gopher", "Hugging_Face", "In-context_learning", "InstructGPT", "Instruction_tuning", "Knowledge_distillation", "LLaMA", "LLaVA", "Llama_(model_family)", "MMLU", "Masked_language_modeling", "Meta_AI", "Mixtral_8x7B", "Mixture_of_experts", "Moonshot_AI", "Next-token_prediction", "OpenAI", "Pretraining", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Scaling_laws", "Self-attention", "Softmax", "T5", "Teacher_forcing", "Tokenization", "Transformer_(architecture)", "ULMFiT", "Word2vec", "Zhipu_AI", "o1", "xAI" ], "categories": [ "Concepts", "Models" ] }, "Layer_normalization": { "title": "Layer normalization", "type": "concept", "words": 1297, "refs": 5, "outbound": [ "BERT", "Dropout", "Feed-forward_network", "GPT-1", "Large_language_model", "RMSNorm", "Self-attention", "Stable_Diffusion", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Architectures", "Concepts" ] }, "Ling_(models)": { "title": "Ling (models)", "type": "model", "words": 394, "refs": 3, "outbound": [ "Ant_Group", "Chinese_AI_labs", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "GLM-4.5", "Huawei", "Hugging_Face", "Instruction_tuning", "Kimi_K2", "Large_language_model", "LongCat", "Meituan", "Mixture_of_experts", "NVIDIA", "Qwen3", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Models", "Mixture-of-experts models", "Open-weight models", "2025 model releases", "Chinese AI labs" ] }, "Liquid_AI": { "title": "Liquid AI", "type": "organization", "words": 968, "refs": 6, "outbound": [ "Anthropic", "Falcon_(models)", "Hugging_Face", "Hybrid_architecture_(LLM)", "Jamba", "LFM_(models)", "LSTM", "Large_language_model", "Mistral_AI", "Mixture_of_experts", "OpenAI", "RWKV", "State-space_model", "Stated_missions_of_AI_labs", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Llama_(model_family)": { "title": "Llama (model family)", "type": "family", "words": 347, "refs": 3, "outbound": [ "Alpaca", "Chinchilla", "Code_Llama", "DeepSeek", "DeepSeek_(model_family)", "GPT-3", "GPT-4", "LLaMA", "Large_language_model", "Llama_2", "Llama_3", "Llama_3.1", "Llama_3.2", "Llama_3.3", "Llama_4", "MMLU", "Meta_AI", "Mistral_(model_family)", "Mistral_AI", "Mixture_of_experts", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)", "Vicuna" ], "categories": [ "Model families", "Meta models", "Open-weight models" ] }, "Llama_2": { "title": "Llama 2", "type": "model", "words": 305, "refs": 3, "outbound": [ "Ai2", "Anthropic", "Code_Llama", "Hugging_Face", "LLaMA", "Llama_(model_family)", "Llama_3", "MMLU", "Meta_AI", "Mistral_AI", "OpenAI", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "Open-weight models", "2023 model releases" ] }, "Llama_3.1": { "title": "Llama 3.1", "type": "model", "words": 354, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "Direct_preference_optimization", "Distributed_training", "GPT-4o", "Hugging_Face", "Instruction_tuning", "Knowledge_distillation", "Large_language_model", "Llama_(model_family)", "Llama_3", "Llama_3.2", "Llama_3.3", "Llama_4", "MMLU", "Meta_AI", "Mixture_of_experts", "NVIDIA", "Pretraining", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "Open-weight models", "2024 model releases" ] }, "Llama_3.2": { "title": "Llama 3.2", "type": "model", "words": 364, "refs": 3, "outbound": [ "Direct_preference_optimization", "Gemma", "Hugging_Face", "Instruction_tuning", "Knowledge_distillation", "Llama_(model_family)", "Llama_3", "Llama_3.1", "Llama_3.3", "Meta_AI", "Qwen2.5", "SmolLM", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "2024 model releases", "Open-weight models", "Multimodal models" ] }, "Llama_3.3": { "title": "Llama 3.3", "type": "model", "words": 364, "refs": 3, "outbound": [ "DeepSeek-V3", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Llama_3", "Llama_3.1", "Llama_3.2", "Llama_4", "MMLU", "Meta_AI", "Qwen2.5", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "Open-weight models", "2024 model releases" ] }, "Llama_3": { "title": "Llama 3", "type": "model", "words": 328, "refs": 3, "outbound": [ "DeepSeek-V3", "Direct_preference_optimization", "Distributed_training", "GPT-4", "Hugging_Face", "LLaMA", "Llama_(model_family)", "Llama_2", "Llama_3.1", "Llama_4", "MMLU", "Meta_AI", "Mixture_of_experts", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "Open-weight models", "2024 model releases" ] }, "Llama_4": { "title": "Llama 4", "type": "model", "words": 403, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek_(model_family)", "Direct_preference_optimization", "GPT-4.5", "Llama_(model_family)", "Llama_3", "MMLU", "Meta_AI", "Mixture_of_experts", "NVIDIA", "Qwen_(model_family)", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "Open-weight models", "Mixture-of-experts models", "2025 model releases" ] }, "Logit_lens": { "title": "Logit lens", "type": "technique", "words": 1590, "refs": 4, "outbound": [ "Activation_steering", "Activation_verbalizer", "Anthropic", "BLOOM", "Circuits_(interpretability)", "EleutherAI", "Eliciting_Latent_Knowledge", "Feed-forward_network", "GPT-2", "GPT-NeoX-20B", "Large_language_model", "Mechanistic_interpretability", "Next-token_prediction", "OPT", "Pythia", "Self-attention", "Sparse_autoencoder", "Superposition_(interpretability)", "Transformer_(architecture)" ], "categories": [ "Concepts", "Alignment and safety" ] }, "LongCat": { "title": "LongCat", "type": "model", "words": 384, "refs": 4, "outbound": [ "Ant_Group", "Chinese_AI_labs", "DeepSeek", "DeepSeek_(model_family)", "Hugging_Face", "Kimi_K3", "MMLU", "Meituan", "Mixture_of_experts", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)", "iFlytek" ], "categories": [ "Chinese AI labs", "Models", "Open-weight models", "Mixture-of-experts models", "2026 model releases" ] }, "MAI-1": { "title": "MAI-1", "type": "model", "words": 562, "refs": 7, "outbound": [ "DALL-E_3", "GPT-4", "Large_language_model", "Megatron-Turing_NLG", "Microsoft_AI", "Mixture_of_experts", "NVIDIA", "OpenAI", "Phi_(model_family)", "Turing-NLG" ], "categories": [ "Microsoft models", "Models", "Mixture-of-experts models", "2025 model releases" ] }, "MMLU": { "title": "MMLU", "type": "benchmark", "words": 333, "refs": 4, "outbound": [ "BERT", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Large_language_model", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Benchmarks", "Concepts" ] }, "MPT_(models)": { "title": "MPT (models)", "type": "model", "words": 293, "refs": 3, "outbound": [ "DBRX", "Databricks", "Falcon_(models)", "FlashAttention", "Instruction_tuning", "LLaMA", "Llama_2", "Mixture_of_experts", "MosaicML", "NVIDIA", "Positional_encoding", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2023 model releases" ] }, "Machines_of_Loving_Grace": { "title": "Machines of Loving Grace", "type": "essay", "words": 604, "refs": 5, "outbound": [ "AI_2027", "Anthropic", "Constitutional_AI", "Mechanistic_interpretability", "Scaling_laws", "Seminal_AI_essays", "Situational_Awareness_(essay)", "The_Most_Important_Century" ], "categories": [ "Essays", "AI governance" ] }, "Magistral": { "title": "Magistral", "type": "model", "words": 380, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "Codestral", "DeepSeek-R1", "Devstral", "Hugging_Face", "Mistral_(model_family)", "Mistral_AI", "Mistral_Large", "Mistral_Medium_3", "Mistral_Small", "Pixtral", "QwQ-32B", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Mistral models", "Reasoning models", "Models", "2025 model releases", "Open-weight models" ] }, "Masked_language_modeling": { "title": "Masked language modeling", "type": "concept", "words": 1503, "refs": 8, "outbound": [ "ALBERT", "BERT", "Cross-entropy_loss", "DistilBERT", "ELECTRA", "Embedding_(machine_learning)", "GPT-1", "Next-token_prediction", "Pretraining", "RoBERTa", "Softmax", "T5", "Transformer_(architecture)", "XLNet" ], "categories": [ "Concepts", "Training methods" ] }, "Mechanistic_interpretability": { "title": "Mechanistic interpretability", "type": "concept", "words": 2790, "refs": 16, "outbound": [ "Activation_steering", "Activation_verbalizer", "Analytic_distillation", "Anthropic", "Attention_sink", "BERT", "CLIP", "Chain-of-thought_prompting", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_3", "Claude_Opus_4.6", "Compositional_generalization", "Constitutional_AI", "EleutherAI", "Eliciting_Latent_Knowledge", "Feed-forward_network", "GPT-2", "GPT-4", "Gemma", "Google_DeepMind", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Logit_lens", "Multi-head_attention", "OLMo", "OpenAI", "Pythia", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Self-attention", "Sparse_autoencoder", "Superposition_(interpretability)", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Alignment and safety", "Concepts" ] }, "Meena": { "title": "Meena", "type": "model", "words": 380, "refs": 3, "outbound": [ "BERT", "GPT-2", "GPT-3", "Gemini_(model_family)", "Google_DeepMind", "LaMDA", "Large_language_model", "OpenAI", "PaLM", "Scaling_laws", "T5", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2020 model releases" ] }, "Megatron-LM": { "title": "Megatron-LM", "type": "model", "words": 371, "refs": 3, "outbound": [ "BERT", "BLOOM", "Distributed_training", "EleutherAI", "GPT-2", "GPT-3", "GPT-NeoX-20B", "Large_language_model", "Megatron-Turing_NLG", "NVIDIA", "Nemotron", "Scaling_laws", "Self-attention", "Transformer_(architecture)" ], "categories": [ "NVIDIA models", "2019 model releases", "Open-source AI", "Models" ] }, "Megatron-Turing_NLG": { "title": "Megatron-Turing NLG", "type": "model", "words": 386, "refs": 3, "outbound": [ "BLOOM", "Chinchilla", "Common_Crawl", "Distributed_training", "EleutherAI", "GPT-3", "Gopher", "Large_language_model", "Megatron-LM", "Microsoft_AI", "NVIDIA", "Nemotron", "OPT", "PaLM", "Scaling_laws", "The_Pile", "Transformer_(architecture)" ], "categories": [ "NVIDIA models", "Models", "2021 model releases" ] }, "Meituan": { "title": "Meituan", "type": "organization", "words": 566, "refs": 5, "outbound": [ "Ant_Group", "Chinese_AI_labs", "DeepSeek", "DeepSeek_(model_family)", "Huawei", "Kimi_(model_family)", "LongCat", "MMLU", "Mixture_of_experts", "Qwen_(model_family)", "Tencent" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Meta_AI": { "title": "Meta AI", "type": "organization", "words": 618, "refs": 5, "outbound": [ "Anthropic", "Google_DeepMind", "LLaMA", "Large_language_model", "Llama_(model_family)", "Llama_2", "Llama_3", "Llama_4", "MMLU", "Movie_Gen", "OPT", "OpenAI", "Reinforcement_learning_from_human_feedback", "Segment_Anything", "Supervised_fine-tuning", "Transformer_(architecture)", "Yann_LeCun" ], "categories": [ "Organizations", "Frontier labs" ] }, "MiMo": { "title": "MiMo", "type": "model", "words": 364, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek_(model_family)", "GLM_(model_family)", "Hugging_Face", "Kimi_(model_family)", "MMLU", "Mistral_7B", "Mistral_AI", "Phi_(model_family)", "Pythia", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)", "Xiaomi" ], "categories": [ "Chinese AI labs", "Models", "Open-weight models", "Reasoning models", "2026 model releases" ] }, "Microsoft_AI": { "title": "Microsoft AI", "type": "organization", "words": 1751, "refs": 25, "outbound": [ "Codex_(2021_model)", "GPT-3", "GPT-4", "Google_DeepMind", "Grok_3", "Large_language_model", "Llama_(model_family)", "MAI-1", "Megatron-LM", "Megatron-Turing_NLG", "Meta_AI", "Mistral_AI", "Mistral_Large", "Mixture_of_experts", "NVIDIA", "OpenAI", "Phi_(model_family)", "Prometheus", "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)", "Safe_Superintelligence", "Stated_missions_of_AI_labs", "Transformer_(architecture)", "Turing-NLG", "xAI" ], "categories": [ "Organizations", "Frontier labs" ] }, "MiniMax-M1": { "title": "MiniMax-M1", "type": "model", "words": 371, "refs": 3, "outbound": [ "DeepSeek-R1", "Gemini_2.5", "Hugging_Face", "Kimi_K2", "MiniMax", "MiniMax-M2", "MiniMax-Text-01", "Mixture_of_experts", "Qwen3", "Reinforcement_learning_with_verifiable_rewards", "Self-attention", "Softmax", "Transformer_(architecture)", "o3" ], "categories": [ "MiniMax models", "Models", "Reasoning models", "Mixture-of-experts models", "Open-weight models", "2025 model releases" ] }, "MiniMax-M2": { "title": "MiniMax-M2", "type": "model", "words": 397, "refs": 3, "outbound": [ "Chinese_AI_labs", "DeepSeek-R1", "DeepSeek-V3", "GLM-4.6", "Hailuo", "Hugging_Face", "Kimi_K2", "Large_language_model", "MiniMax", "MiniMax-M1", "MiniMax-Text-01", "Mixture_of_experts", "Qwen3", "Transformer_(architecture)" ], "categories": [ "MiniMax models", "Models", "2025 model releases", "Mixture-of-experts models", "Open-weight models", "Reasoning models" ] }, "MiniMax-Text-01": { "title": "MiniMax-Text-01", "type": "model", "words": 372, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "DeepSeek-V3", "GPT-4o", "Gemini_1.5", "Hailuo", "Hugging_Face", "Instruction_tuning", "Large_language_model", "MiniMax", "MiniMax-M1", "MiniMax-M2", "Mixture_of_experts", "Qwen2.5", "Softmax", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "MiniMax models", "Models", "Mixture-of-experts models", "Open-weight models", "2025 model releases" ] }, "MiniMax": { "title": "MiniMax", "type": "organization", "words": 525, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek_(model_family)", "Hailuo", "Kimi_(model_family)", "MMLU", "MiniMax-M1", "MiniMax-M2", "MiniMax-Text-01", "Mixture_of_experts", "Moonshot_AI", "Qwen_(model_family)", "SenseTime", "Tencent", "Test-time_compute" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Mistral_(model_family)": { "title": "Mistral (model family)", "type": "family", "words": 292, "refs": 3, "outbound": [ "Anthropic", "Codestral", "DeepSeek_(model_family)", "Devstral", "GPT-3.5", "Hugging_Face", "Llama_(model_family)", "MMLU", "Magistral", "Meta_AI", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mistral_Medium_3", "Mistral_Small", "Mixtral_8x22B", "Mixtral_8x7B", "Mixture_of_experts", "OpenAI", "Pixtral", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Model families", "Mistral models", "Open-weight models" ] }, "Mistral_7B": { "title": "Mistral 7B", "type": "model", "words": 409, "refs": 3, "outbound": [ "Alpaca", "Chinchilla", "Code_Llama", "Codestral", "Hugging_Face", "LLaMA", "Llama_2", "MMLU", "Mistral_(model_family)", "Mistral_AI", "Mistral_Large", "Mistral_Small", "Mixtral_8x7B", "Mixture_of_experts", "Sliding-window_attention", "Transformer_(architecture)", "Vicuna" ], "categories": [ "Mistral models", "Models", "2023 model releases", "Open-weight models" ] }, "Mistral_AI": { "title": "Mistral AI", "type": "organization", "words": 569, "refs": 4, "outbound": [ "Anthropic", "Codestral", "Cohere", "Command_(models)", "Google_DeepMind", "Instruction_tuning", "LLaMA", "Large_language_model", "MMLU", "Magistral", "Meta_AI", "Mistral_(model_family)", "Mistral_7B", "Mistral_Large", "Mistral_Medium_3", "Mistral_Small", "Mixtral_8x7B", "Mixture_of_experts", "OpenAI", "Pixtral", "Reinforcement_learning_from_human_feedback", "Sliding-window_attention", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] }, "Mistral_Large": { "title": "Mistral Large", "type": "model", "words": 365, "refs": 3, "outbound": [ "Codestral", "GPT-4", "GPT-4o", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_3.1", "MMLU", "Magistral", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Medium_3", "Mistral_Small", "Mixtral_8x7B", "Mixture_of_experts", "Pixtral", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "2024 model releases", "Open-weight models" ] }, "Mistral_Medium_3": { "title": "Mistral Medium 3", "type": "model", "words": 361, "refs": 3, "outbound": [ "Claude_3.7_Sonnet", "Codestral", "Devstral", "Instruction_tuning", "Large_language_model", "Llama_4", "Magistral", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mistral_Small", "Mixtral_8x7B", "Pixtral", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "2025 model releases" ] }, "Mistral_Small": { "title": "Mistral Small", "type": "model", "words": 366, "refs": 3, "outbound": [ "Codestral", "Devstral", "Gemma", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_3.3", "MMLU", "Magistral", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mistral_Medium_3", "Mixtral_8x7B", "Mixture_of_experts", "Pixtral", "Qwen2.5", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "Open-weight models", "2024 model releases", "2025 model releases" ] }, "Mixtral_8x22B": { "title": "Mixtral 8x22B", "type": "model", "words": 404, "refs": 3, "outbound": [ "Codestral", "Hugging_Face", "Large_language_model", "Llama_3", "MMLU", "Meta_AI", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mistral_Small", "Mixtral_8x7B", "Mixture_of_experts", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "Mixture-of-experts models", "Open-weight models", "2024 model releases" ] }, "Mixtral_8x7B": { "title": "Mixtral 8x7B", "type": "model", "words": 392, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek-V2", "Direct_preference_optimization", "GPT-3.5", "GPT-4", "Hugging_Face", "Llama_2", "Llama_4", "MMLU", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mixtral_8x22B", "Mixture_of_experts", "PagedAttention", "Qwen_(model_family)", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "Open-weight models", "Mixture-of-experts models", "2023 model releases" ] }, "Mixture_of_experts": { "title": "Mixture of experts", "type": "concept", "words": 2951, "refs": 30, "outbound": [ "Ai2", "Ant_Group", "Anthropic", "Arcee_AI", "ByteDance", "Chinchilla", "Chinese_AI_labs", "Claude_(model_family)", "DBRX", "Databricks", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek-V3.1", "DeepSeek_(model_family)", "Distributed_training", "Feed-forward_network", "GLM-4.5", "GLM-5", "GLM_(model_family)", "GPT-3", "GPT-4", "GPT-5", "Gemini_1.5", "Google_DeepMind", "Grok-1", "Hunyuan", "Hybrid_architecture_(LLM)", "Instruction_tuning", "Kimi_K2", "Kimi_K3", "Knowledge_distillation", "LSTM", "Large_language_model", "Ling_(models)", "Llama_(model_family)", "Llama_4", "LongCat", "MMLU", "Meituan", "Meta_AI", "MiniMax", "MiniMax-Text-01", "Mistral_AI", "Mixtral_8x22B", "Mixtral_8x7B", "Moonshot_AI", "Multi-head_latent_attention", "NVIDIA", "OLMo", "OpenAI", "PaLM", "Qwen3", "Qwen3-Max", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Seed-OSS", "Self-attention", "Sliding-window_attention", "Softmax", "Sparse_attention", "Tencent", "Test-time_compute", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Mini", "Vision_Transformer", "Zhipu_AI", "gpt-oss", "xAI" ], "categories": [ "Architectures", "Concepts", "Training methods" ] }, "Molmo": { "title": "Molmo", "type": "model", "words": 367, "refs": 3, "outbound": [ "Ai2", "CLIP", "GPT-4", "GPT-4o", "Gemini_1.5", "Hugging_Face", "Instruction_tuning", "InternVL", "Large_language_model", "Mixture_of_experts", "OLMo", "OpenAI", "PaliGemma", "Pixtral", "Qwen2", "Qwen2.5-VL", "Transformer_(architecture)", "Tulu_3" ], "categories": [ "Models", "Multimodal models", "Open-weight models", "2024 model releases" ] }, "Moondream": { "title": "Moondream", "type": "organization", "words": 551, "refs": 5, "outbound": [ "Apple_foundation_models", "EleutherAI", "GPT-4o", "Hugging_Face", "Large_language_model", "Mistral_AI", "Mixture_of_experts", "Moondream_(model_family)", "Phi_(model_family)", "Stability_AI", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Moondream_(model_family)": { "title": "Moondream (model family)", "type": "family", "words": 340, "refs": 3, "outbound": [ "Flamingo", "GPT-4", "Gemma", "Hugging_Face", "Large_language_model", "Mixture_of_experts", "Moondream", "Qwen_(model_family)", "Transformer_(architecture)" ], "categories": [ "Model families", "Vision-language models", "Open-weight models" ] }, "Moonshot_AI": { "title": "Moonshot AI", "type": "organization", "words": 526, "refs": 6, "outbound": [ "Chinese_AI_labs", "DeepSeek", "GLM_(model_family)", "Hugging_Face", "Kimi_(model_family)", "Kimi_K2", "Kimi_K3", "Kimi_k1.5", "Large_language_model", "MMLU", "Mixture_of_experts", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)", "Zhipu_AI", "o1" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "MosaicML": { "title": "MosaicML", "type": "organization", "words": 613, "refs": 4, "outbound": [ "Adam_(optimizer)", "DBRX", "Databricks", "Falcon_(models)", "FlashAttention", "LLaMA", "Llama_2", "MPT_(models)", "Mixture_of_experts", "Positional_encoding" ], "categories": [ "Open-source AI", "Frontier labs" ] }, "Movie_Gen": { "title": "Movie Gen", "type": "model", "words": 432, "refs": 3, "outbound": [ "DALL-E", "Hailuo", "HunyuanVideo", "Imagen", "LLaMA", "Llama_(model_family)", "Llama_3", "Meta_AI", "MiniMax", "OPT", "OpenAI", "Segment_Anything", "Sora", "Stable_Diffusion", "Tencent", "Transformer_(architecture)", "Veo" ], "categories": [ "Meta models", "Models", "Multimodal models", "2024 model releases" ] }, "Multi-head_attention": { "title": "Multi-head attention", "type": "technique", "words": 2700, "refs": 21, "outbound": [ "Anthropic", "Attention_sink", "BERT", "CLIP", "Circuits_(interpretability)", "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "Diffusion_language_model", "FlashAttention", "GPT-2", "GPT-3", "Gemini_1.0", "Gemma", "Hybrid_architecture_(LLM)", "In-context_learning", "Kimi_K2", "LLaMA", "Large_language_model", "Llama_2", "Llama_3", "Llama_3.1", "Mechanistic_interpretability", "Megatron-LM", "MiniMax-Text-01", "Mistral_7B", "Mistral_AI", "Mixtral_8x7B", "Mixture_of_experts", "Moonshot_AI", "Multi-head_latent_attention", "OpenAI", "PaLM", "PagedAttention", "Qwen2", "Qwen3", "Self-attention", "Sliding-window_attention", "Softmax", "Sparse_attention", "StarCoder", "State-space_model", "T5", "Transformer_(architecture)", "Vision_Transformer", "Whisper", "gpt-oss" ], "categories": [ "Architectures", "Concepts", "Inference" ] }, "Multi-head_latent_attention": { "title": "Multi-head latent attention", "type": "technique", "words": 1814, "refs": 9, "outbound": [ "DeepSeek", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "DeepSeek-V3.1", "DeepSeek_(model_family)", "FlashAttention", "Hybrid_architecture_(LLM)", "Kimi_K2", "Large_language_model", "Llama_2", "LongCat", "Meituan", "Mixture_of_experts", "Moonshot_AI", "Multi-head_attention", "NVIDIA", "PagedAttention", "Positional_encoding", "Qwen2.5", "Self-attention", "Sliding-window_attention", "Sparse_attention", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Architectures", "Inference", "Concepts" ] }, "NETtalk": { "title": "NETtalk", "type": "model", "words": 1033, "refs": 10, "outbound": [ "Backpropagation", "Boltzmann_machine", "David_Rumelhart", "Feed-forward_network", "Geoffrey_Hinton", "Hopfield_network", "Large_language_model", "Perceptron" ], "categories": [ "Models", "Pre-transformer systems" ] }, "NVIDIA": { "title": "NVIDIA", "type": "organization", "words": 632, "refs": 5, "outbound": [ "Amazon_Nova", "Anthropic", "Distributed_training", "GPT-3", "GPT-4", "Google_DeepMind", "Hugging_Face", "Large_language_model", "MMLU", "Megatron-LM", "Megatron-Turing_NLG", "Meta_AI", "Nemotron", "OpenAI", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)", "xAI" ], "categories": [ "Organizations", "Hardware and compute" ] }, "Ndea": { "title": "Ndea", "type": "organization", "words": 1036, "refs": 8, "outbound": [ "Compositional_generalization", "François_Chollet", "Google_DeepMind", "Keras", "Large_language_model", "MMLU", "OpenAI", "Safe_Superintelligence", "Scaling_laws", "Stated_missions_of_AI_labs", "Test-time_compute", "Transformer_(architecture)", "o3" ], "categories": [ "Organizations", "Frontier labs" ] }, "Nemotron": { "title": "Nemotron", "type": "model", "words": 355, "refs": 3, "outbound": [ "Direct_preference_optimization", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Llama_3.1", "Megatron-LM", "Megatron-Turing_NLG", "NVIDIA", "Reinforcement_learning_from_human_feedback", "Reward_model", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "NVIDIA models", "Models", "Open-weight models", "2024 model releases" ] }, "Next-token_prediction": { "title": "Next-token prediction", "type": "concept", "words": 1441, "refs": 10, "outbound": [ "BERT", "Backpropagation", "Chinchilla", "Cross-entropy_loss", "DeepSeek-V3", "Diffusion_language_model", "Direct_preference_optimization", "GPT-3", "GPT_(model_family)", "In-context_learning", "Large_language_model", "Masked_language_modeling", "Pretraining", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Self-attention", "Seq2seq", "Softmax", "Teacher_forcing", "Transformer_(architecture)" ], "categories": [ "Concepts", "Training methods" ] }, "No_Fire_Alarm_for_Artificial_General_Intelligence": { "title": "There's No Fire Alarm for Artificial General Intelligence", "type": "essay", "words": 698, "refs": 5, "outbound": [ "AI_2027", "AlphaGo", "DeepSeek_R1_release_shock", "GPT-3", "Large_language_model", "Seminal_AI_essays", "Situational_Awareness_(essay)", "The_Most_Important_Century", "What_Failure_Looks_Like" ], "categories": [ "Essays", "Alignment and safety" ] }, "Nous_Research": { "title": "Nous Research", "type": "organization", "words": 771, "refs": 6, "outbound": [ "Anthropic", "Distributed_training", "EleutherAI", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Llama_3.1", "Meta_AI", "Mistral_AI", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Stated_missions_of_AI_labs", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "OLMo": { "title": "OLMo", "type": "model", "words": 380, "refs": 3, "outbound": [ "Ai2", "Anthropic", "Chain-of-thought_prompting", "Chinese_AI_labs", "Direct_preference_optimization", "EleutherAI", "Gemma", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Llama_3.1", "MMLU", "Molmo", "OpenAI", "Pythia", "Reinforcement_learning_with_verifiable_rewards", "Scaling_laws", "Supervised_fine-tuning", "Transformer_(architecture)", "Tulu_3" ], "categories": [ "Models", "Open-weight models", "2024 model releases" ] }, "OPT": { "title": "OPT", "type": "model", "words": 369, "refs": 3, "outbound": [ "BLOOM", "EleutherAI", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Hugging_Face", "LLaMA", "Large_language_model", "Llama_(model_family)", "Meta_AI", "NVIDIA", "OLMo", "OpenAI", "Pythia", "RoBERTa", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Meta models", "Models", "2022 model releases", "Open-weight models" ] }, "OpenAI": { "title": "OpenAI", "type": "organization", "words": 769, "refs": 6, "outbound": [ "AI_and_Compute", "BERT", "DALL-E", "GPT-2", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Ilya_Sutskever", "In-context_learning", "Large_language_model", "MMLU", "Microsoft_AI", "Reinforcement_learning_from_human_feedback", "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)", "Sora", "Transformer_(architecture)", "Whisper", "gpt-oss" ], "categories": [ "Organizations", "Frontier labs" ] }, "PaLM": { "title": "PaLM", "type": "model", "words": 331, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "Chinchilla", "Distributed_training", "Flan-T5", "GPT-3", "Gemini_(model_family)", "Gemini_1.0", "Google_DeepMind", "Gopher", "Instruction_tuning", "LaMDA", "Large_language_model", "MMLU", "Megatron-Turing_NLG", "PaLM_2", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2022 model releases" ] }, "PaLM_2": { "title": "PaLM 2", "type": "model", "words": 357, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "Chinchilla", "GPT-4", "Gemini_(model_family)", "Gemini_1.0", "Google_DeepMind", "Instruction_tuning", "LLaMA", "LaMDA", "Large_language_model", "MMLU", "PaLM", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2023 model releases" ] }, "PagedAttention": { "title": "PagedAttention", "type": "technique", "words": 2685, "refs": 16, "outbound": [ "Attention_sink", "DeepSeek-R1", "DeepSeek-V2", "DeepSeek-V3", "Diffusion_language_model", "FlashAttention", "Gemma", "Hugging_Face", "Hybrid_architecture_(LLM)", "Kimi_(model_family)", "LLaMA", "Large_language_model", "Llama_3.1", "Meta_AI", "Mistral_7B", "Mistral_AI", "Mixtral_8x7B", "Mixture_of_experts", "Moonshot_AI", "Multi-head_attention", "Multi-head_latent_attention", "NVIDIA", "OPT", "OpenAI", "Qwen3", "Self-attention", "Sliding-window_attention", "Sparse_attention", "State-space_model", "Test-time_compute", "Transformer_(architecture)", "Vicuna", "Vision_Transformer", "gpt-oss" ], "categories": [ "Architectures", "Inference", "Concepts" ] }, "PaliGemma": { "title": "PaliGemma", "type": "model", "words": 352, "refs": 3, "outbound": [ "Gemma", "Google_DeepMind", "Hugging_Face", "InternVL", "Llama_3.2", "Molmo", "PaLM", "Pixtral", "Qwen2.5-VL", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Google models", "Models", "Multimodal models", "Open-weight models", "2024 model releases" ] }, "Paul_Werbos": { "title": "Paul Werbos", "type": "person", "words": 1439, "refs": 14, "outbound": [ "Andrew_Ng", "Backpropagation", "David_Rumelhart", "Feed-forward_network", "Frank_Rosenblatt", "LSTM", "Large_language_model", "Perceptron", "Reinforcement_learning_from_human_feedback", "Reward_model", "Richard_Sutton", "Seq2seq", "Transformer_(architecture)" ], "categories": [ "People", "Computer science pioneers" ] }, "Perceptron": { "title": "Perceptron", "type": "concept", "words": 1363, "refs": 14, "outbound": [ "Backpropagation", "Boltzmann_machine", "Feed-forward_network", "Frank_Rosenblatt", "Hopfield_network", "Softmax", "Walter_Pitts", "Warren_McCulloch" ], "categories": [ "Concepts", "Architectures" ] }, "Phi_(model_family)": { "title": "Phi (model family)", "type": "family", "words": 316, "refs": 3, "outbound": [ "Gemma", "Hugging_Face", "MMLU", "MiMo", "Microsoft_AI", "Mixture_of_experts", "OLMo", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Model families", "Open-weight models", "Microsoft models" ] }, "Pixtral": { "title": "Pixtral", "type": "model", "words": 354, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "GPT-4o", "Gemini_1.5", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_3.2", "Mistral_(model_family)", "Mistral_7B", "Mistral_AI", "Mistral_Large", "Mixtral_8x7B", "Molmo", "Qwen2", "Transformer_(architecture)" ], "categories": [ "Mistral models", "Models", "Multimodal models", "Open-weight models", "2024 model releases" ] }, "Positional_encoding": { "title": "Positional encoding", "type": "concept", "words": 1590, "refs": 14, "outbound": [ "BERT", "BLOOM", "Embedding_(machine_learning)", "Falcon_(models)", "GPT-1", "GPT-3", "Gemma", "LLaMA", "Llama_(model_family)", "Llama_3", "MPT_(models)", "Mistral_(model_family)", "Multi-head_attention", "OLMo", "Pythia", "Qwen_(model_family)", "Self-attention", "Sliding-window_attention", "T5", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "Architectures", "Concepts" ] }, "Pretraining": { "title": "Pretraining", "type": "concept", "words": 1444, "refs": 13, "outbound": [ "Adam_(optimizer)", "BERT", "Backpropagation", "Chinchilla", "Common_Crawl", "Distributed_training", "ELMo", "GPT-1", "GPT-3", "InstructGPT", "Instruction_tuning", "LLaMA", "LSTM", "Large_language_model", "Masked_language_modeling", "Next-token_prediction", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Supervised_fine-tuning", "Teacher_forcing", "The_Pile", "Tokenization", "Transformer_(architecture)", "ULMFiT", "Word2vec" ], "categories": [ "Training methods", "Concepts" ] }, "Prometheus": { "title": "Prometheus", "type": "organization", "words": 1369, "refs": 8, "outbound": [ "Anthropic", "Google_DeepMind", "Large_language_model", "Meta_AI", "Microsoft_AI", "NVIDIA", "Ndea", "OpenAI", "Pretraining", "Safe_Superintelligence", "Stated_missions_of_AI_labs", "Thinking_Machines_Lab", "xAI" ], "categories": [ "Organizations", "Frontier labs" ] }, "Pythia": { "title": "Pythia", "type": "model", "words": 639, "refs": 9, "outbound": [ "Ai2", "BLOOM", "Chinchilla", "Databricks", "EleutherAI", "GPT-3", "GPT-J", "GPT-NeoX-20B", "Hugging_Face", "Mechanistic_interpretability", "Meta_AI", "OLMo", "OPT", "Scaling_laws", "Sparse_autoencoder", "Superposition_(interpretability)", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Open-source AI", "2023 model releases" ] }, "QwQ-32B": { "title": "QwQ-32B", "type": "model", "words": 351, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "DeepSeek-R1", "DeepSeek_R1_release_shock", "Hugging_Face", "Kimi_k1.5", "OpenAI", "Qwen2.5", "Qwen3", "Qwen_(model_family)", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Alibaba models", "Models", "Reasoning models", "Open-weight models", "2025 model releases" ] }, "Qwen2.5-Max": { "title": "Qwen2.5-Max", "type": "model", "words": 372, "refs": 3, "outbound": [ "Claude_3.5_Sonnet", "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "DeepSeek_R1_release_shock", "GPT-4o", "Instruction_tuning", "Large_language_model", "Mixture_of_experts", "QwQ-32B", "Qwen2.5", "Qwen2.5-VL", "Qwen3", "Qwen3-Max", "Qwen_(model_family)", "Qwen_team", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "Mixture-of-experts models", "2025 model releases" ] }, "Qwen2.5-VL": { "title": "Qwen2.5-VL", "type": "model", "words": 350, "refs": 3, "outbound": [ "GPT-4o", "Hugging_Face", "Instruction_tuning", "InternVL", "Large_language_model", "Llama_3.2", "Molmo", "PaliGemma", "Pixtral", "Qwen2.5", "Qwen2.5-Max", "Qwen3", "Qwen_(model_family)", "Qwen_team", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "Multimodal models", "Open-weight models", "2025 model releases", "Chinese AI labs" ] }, "Qwen2.5": { "title": "Qwen2.5", "type": "model", "words": 331, "refs": 3, "outbound": [ "DeepSeek", "DeepSeek-R1", "Hugging_Face", "Instruction_tuning", "Llama_3.1", "Mixture_of_experts", "QwQ-32B", "Qwen2", "Qwen2.5-Max", "Qwen2.5-VL", "Qwen3", "Qwen_(model_family)", "Qwen_team", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "Open-weight models", "2024 model releases" ] }, "Qwen2": { "title": "Qwen2", "type": "model", "words": 355, "refs": 3, "outbound": [ "DeepSeek-V2", "Direct_preference_optimization", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_3", "MMLU", "Meta_AI", "Mixtral_8x7B", "Mixture_of_experts", "Qwen2.5", "Qwen2.5-VL", "Qwen3", "Qwen_(model_family)", "Qwen_1", "Qwen_team", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "Open-weight models", "Mixture-of-experts models", "2024 model releases" ] }, "Qwen3-Coder-Next": { "title": "Qwen3-Coder-Next", "type": "model", "words": 381, "refs": 3, "outbound": [ "Claude_Opus_4.5", "Code_Llama", "Codestral", "DeepSeek-Coder-V2", "Devstral", "GLM-4.6", "GPT-5.1", "Hugging_Face", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Qwen3", "Qwen3-Coder", "Qwen_(model_family)", "Qwen_team", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "2026 model releases", "Open-weight models", "Mixture-of-experts models", "Models" ] }, "Qwen3-Coder": { "title": "Qwen3-Coder", "type": "model", "words": 370, "refs": 3, "outbound": [ "Code_Llama", "Codestral", "DeepSeek-Coder-V2", "Devstral", "GLM-4.5", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "Qwen2.5", "Qwen3", "Qwen3-Coder-Next", "Qwen_(model_family)", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "Open-weight models", "Mixture-of-experts models", "2025 model releases", "Chinese AI labs" ] }, "Qwen3-Max": { "title": "Qwen3-Max", "type": "model", "words": 380, "refs": 3, "outbound": [ "DeepSeek-V3", "GLM-4.5", "GPT-5", "Instruction_tuning", "Kimi_K2", "Mixture_of_experts", "Qwen2.5", "Qwen2.5-Max", "Qwen3", "Qwen3-Coder", "Qwen_(model_family)", "Qwen_team", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Chinese AI labs", "Models", "Mixture-of-experts models", "2025 model releases" ] }, "Qwen3": { "title": "Qwen3", "type": "model", "words": 328, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "DeepSeek-R1", "Gemini_2.5", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_4", "MMLU", "Mixture_of_experts", "QwQ-32B", "Qwen2.5", "Qwen3-Coder", "Qwen3-Max", "Qwen_(model_family)", "Qwen_team", "Reinforcement_learning_from_human_feedback", "Test-time_compute", "Transformer_(architecture)", "o1" ], "categories": [ "Alibaba models", "Models", "Open-weight models", "Reasoning models", "Mixture-of-experts models", "2025 model releases" ] }, "Qwen_(model_family)": { "title": "Qwen (model family)", "type": "family", "words": 286, "refs": 3, "outbound": [ "DeepSeek", "DeepSeek_(model_family)", "Hugging_Face", "Kimi_(model_family)", "Llama_(model_family)", "MMLU", "Mistral_(model_family)", "Mixture_of_experts", "QwQ-32B", "Qwen2", "Qwen2.5", "Qwen2.5-Max", "Qwen2.5-VL", "Qwen3", "Qwen3-Coder", "Qwen3-Coder-Next", "Qwen3-Max", "Qwen_1", "Qwen_team", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Model families", "Alibaba models", "Open-weight models" ] }, "Qwen_1": { "title": "Qwen 1", "type": "model", "words": 411, "refs": 3, "outbound": [ "Baichuan", "ChatGLM", "Chinese_AI_labs", "Hugging_Face", "Instruction_tuning", "InternLM", "LLaMA", "Large_language_model", "Llama_2", "MMLU", "Qwen2", "Qwen2.5", "Qwen3", "Qwen_(model_family)", "Qwen_team", "RMSNorm", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Alibaba models", "Models", "2023 model releases", "Open-weight models", "Chinese AI labs" ] }, "Qwen_team": { "title": "Qwen team", "type": "organization", "words": 522, "refs": 5, "outbound": [ "Chinese_AI_labs", "DeepSeek", "DeepSeek-R1", "DeepSeek_(model_family)", "GLM_(model_family)", "Hugging_Face", "Kimi_(model_family)", "Llama_(model_family)", "MMLU", "Meta_AI", "QwQ-32B", "Qwen2", "Qwen2.5", "Qwen2.5-Max", "Qwen2.5-VL", "Qwen3", "Qwen3-Coder", "Qwen3-Max", "Qwen_(model_family)", "Qwen_1", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning" ], "categories": [ "Organizations", "Chinese AI labs", "Frontier labs" ] }, "RMSNorm": { "title": "RMSNorm", "type": "concept", "words": 1149, "refs": 5, "outbound": [ "BERT", "DeepSeek_(model_family)", "Feed-forward_network", "Gemma", "Google_DeepMind", "Gopher", "LLaMA", "Large_language_model", "Layer_normalization", "Meta_AI", "Mistral_7B", "Qwen_(model_family)", "Self-attention", "T5", "Transformer_(architecture)" ], "categories": [ "Architectures", "Concepts" ] }, "RWKV": { "title": "RWKV", "type": "model", "words": 329, "refs": 3, "outbound": [ "EleutherAI", "Jamba", "Kimi_K3", "LSTM", "MMLU", "Moonshot_AI", "Self-attention", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Architectures" ] }, "Reinforcement_learning_from_human_feedback": { "title": "Reinforcement learning from human feedback", "type": "technique", "words": 336, "refs": 3, "outbound": [ "Direct_preference_optimization", "GPT-3", "GPT-3.5", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "InstructGPT", "Large_language_model", "MMLU", "OpenAI", "Reward_model", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Post-training", "Training methods", "Alignment and safety" ] }, "Reinforcement_learning_with_verifiable_rewards": { "title": "Reinforcement learning with verifiable rewards", "type": "technique", "words": 326, "refs": 3, "outbound": [ "Claude_(model_family)", "DeepSeek", "DeepSeek-R1", "DeepSeek_(model_family)", "Direct_preference_optimization", "Gemini_(model_family)", "Large_language_model", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Reward_model", "Supervised_fine-tuning", "Tulu_3", "o1" ], "categories": [ "Post-training", "Training methods", "Reasoning models" ] }, "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)": { "title": "Removal of Sam Altman from OpenAI (annotated edition)", "type": "incident", "words": 7492, "refs": 106, "outbound": [ "Anthropic", "Cohere", "DeepSeek_R1_release_shock", "GPT-1", "GPT-2", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Ilya_Sutskever", "Microsoft_AI", "OpenAI", "xAI" ], "categories": [ "Incidents and controversies", "AI governance" ] }, "Reward_model": { "title": "Reward model", "type": "concept", "words": 361, "refs": 4, "outbound": [ "Claude_(model_family)", "Constitutional_AI", "DeepSeek-R1", "Direct_preference_optimization", "GPT-4", "InstructGPT", "OpenAI", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Concepts", "Post-training" ] }, "Richard_Sutton": { "title": "Richard Sutton", "type": "person", "words": 2439, "refs": 25, "outbound": [ "AlexNet", "Alex_Krizhevsky", "AlphaGo", "AlphaZero", "Arthur_Samuel", "Backpropagation", "Chain-of-thought_prompting", "David_Silver", "DeepSeek-R1", "GPT-4", "Google_DeepMind", "Ilya_Sutskever", "ImageNet", "InstructGPT", "Large_language_model", "Next-token_prediction", "OpenAI", "Pretraining", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Reward_model", "Scaling_laws", "Supervised_fine-tuning", "TD-Gammon", "Test-time_compute", "The_Bitter_Lesson", "Transformer_(architecture)", "o1" ], "categories": [ "People", "Machine learning researchers" ] }, "RoBERTa": { "title": "RoBERTa", "type": "model", "words": 408, "refs": 3, "outbound": [ "ALBERT", "BERT", "DistilBERT", "ELECTRA", "GPT-1", "GPT-2", "GPT-3", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "Masked_language_modeling", "Meta_AI", "Pretraining", "Scaling_laws", "Transformer_(architecture)", "XLNet" ], "categories": [ "Meta models", "Models", "2019 model releases", "Open-weight models", "Open-source AI" ] }, "Safe_Superintelligence": { "title": "Safe Superintelligence", "type": "organization", "words": 1960, "refs": 20, "outbound": [ "AlexNet", "Anthropic", "DeepSeek", "GPT-3", "GPT-4", "GPT_(model_family)", "Google_DeepMind", "Ilya_Sutskever", "Large_language_model", "Meta_AI", "Mistral_AI", "NVIDIA", "OpenAI", "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)", "Scaling_laws", "Stated_missions_of_AI_labs", "Test-time_compute", "Transformer_(architecture)", "xAI" ], "categories": [ "Organizations", "Frontier labs" ] }, "Scaling_laws": { "title": "Scaling laws", "type": "concept", "words": 408, "refs": 7, "outbound": [ "AI_and_Compute", "Anthropic", "Chinchilla", "DeepSeek", "GPT-3", "Google_DeepMind", "Gopher", "Large_language_model", "Llama_(model_family)", "MMLU", "Meta_AI", "OpenAI", "Richard_Sutton", "Situational_Awareness_(essay)", "Test-time_compute", "The_Bitter_Lesson", "The_Scaling_Hypothesis", "Transformer_(architecture)", "o1" ], "categories": [ "Concepts", "Training methods" ] }, "Seed-OSS": { "title": "Seed-OSS", "type": "model", "words": 335, "refs": 3, "outbound": [ "ByteDance", "Chinese_AI_labs", "DeepSeek-R1", "Doubao", "GLM-4.5", "Hugging_Face", "Instruction_tuning", "Kimi_K2", "Large_language_model", "Mixture_of_experts", "OpenAI", "Qwen3", "Seedream", "Test-time_compute", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "ByteDance models", "Chinese AI labs", "Open-weight models", "2025 model releases", "Models" ] }, "Seedream": { "title": "Seedream", "type": "model", "words": 339, "refs": 4, "outbound": [ "ByteDance", "Chinese_AI_labs", "DALL-E", "Doubao", "Hailuo", "HunyuanVideo", "Imagen", "Seed-OSS", "Sora", "Stable_Diffusion", "Transformer_(architecture)", "Veo" ], "categories": [ "Chinese AI labs", "Models", "Multimodal models", "2025 model releases" ] }, "Segment_Anything": { "title": "Segment Anything", "type": "model", "words": 384, "refs": 3, "outbound": [ "GPT-3.5", "GPT-4", "Hugging_Face", "LLaMA", "Large_language_model", "Meta_AI", "Mixtral_8x7B", "Movie_Gen", "Reinforcement_learning_from_human_feedback", "Self-attention", "Stable_Diffusion", "Transformer_(architecture)", "Vision_Transformer", "Whisper" ], "categories": [ "Meta models", "Models", "2023 model releases", "Open-weight models" ] }, "Self-attention": { "title": "Self-attention", "type": "concept", "words": 3065, "refs": 25, "outbound": [ "Attention_sink", "BERT", "CLIP", "Claude_(model_family)", "DeepSeek", "DeepSeek-V2", "DeepSeek-V3", "Diffusion_language_model", "Feed-forward_network", "FlashAttention", "GPT-2", "GPT-3", "GPT_(model_family)", "Gemini_1.5", "Hybrid_architecture_(LLM)", "Kimi_K2", "LSTM", "Large_language_model", "Layer_normalization", "Llama_(model_family)", "Llama_2", "Llama_3", "Llama_3.1", "MMLU", "MiniMax-Text-01", "Mistral_7B", "Mixtral_8x7B", "Mixture_of_experts", "Multi-head_attention", "Multi-head_latent_attention", "OpenAI", "PagedAttention", "Positional_encoding", "Scaling_laws", "Sliding-window_attention", "Softmax", "Sparse_attention", "Stable_Diffusion", "State-space_model", "T5", "Transformer_(architecture)", "Vision_Transformer", "Whisper", "gpt-oss" ], "categories": [ "Architectures", "Concepts" ] }, "Seminal_AI_essays": { "title": "Seminal AI essays", "type": "reference", "words": 521, "refs": 3, "outbound": [ "AI_2027", "AI_and_Compute", "GPT-3", "Large_language_model", "Machines_of_Loving_Grace", "No_Fire_Alarm_for_Artificial_General_Intelligence", "Scaling_laws", "Simulators_(essay)", "Situational_Awareness_(essay)", "Software_2.0", "Stated_missions_of_AI_labs", "The_Bitter_Lesson", "The_Illustrated_Transformer", "The_Most_Important_Century", "The_Scaling_Hypothesis", "The_Unreasonable_Effectiveness_of_Recurrent_Neural_Networks", "Transformer_(architecture)", "What_Failure_Looks_Like" ], "categories": [ "Essays", "Reference" ] }, "SenseNova": { "title": "SenseNova", "type": "model", "words": 386, "refs": 4, "outbound": [ "Baidu", "Chinese_AI_labs", "Doubao", "ERNIE_Bot", "GPT-4", "GPT-4o", "Hunyuan", "InternLM", "Large_language_model", "Mixture_of_experts", "SenseTime", "Shanghai_AI_Laboratory", "Transformer_(architecture)", "iFlytek_Spark" ], "categories": [ "SenseTime models", "Models", "Multimodal models", "2023 model releases", "Chinese AI labs" ] }, "SenseTime": { "title": "SenseTime", "type": "organization", "words": 592, "refs": 6, "outbound": [ "Baidu", "Chinese_AI_labs", "DeepSeek_(model_family)", "Huawei", "Large_language_model", "MMLU", "MiniMax", "Qwen_(model_family)", "SenseNova", "Shanghai_AI_Laboratory", "iFlytek" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Seq2seq": { "title": "Seq2seq", "type": "concept", "words": 1665, "refs": 14, "outbound": [ "Backpropagation", "Cross-entropy_loss", "Dropout", "Feed-forward_network", "Google_DeepMind", "LSTM", "Next-token_prediction", "Pretraining", "Self-attention", "T5", "Teacher_forcing", "Transformer_(architecture)", "Whisper" ], "categories": [ "Architectures", "Concepts" ] }, "Shanghai_AI_Laboratory": { "title": "Shanghai AI Laboratory", "type": "organization", "words": 557, "refs": 6, "outbound": [ "Ai2", "Chinese_AI_labs", "EleutherAI", "GPT_(model_family)", "Gemini_(model_family)", "Hugging_Face", "InternLM", "InternVL", "MMLU", "Qwen_team", "SenseTime" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Simulators_(essay)": { "title": "Simulators (essay)", "type": "essay", "words": 740, "refs": 5, "outbound": [ "Chain-of-thought_prompting", "Eliciting_Latent_Knowledge", "GPT-3", "In-context_learning", "Large_language_model", "Next-token_prediction", "Reinforcement_learning_from_human_feedback", "Seminal_AI_essays", "The_Unreasonable_Effectiveness_of_Recurrent_Neural_Networks", "What_Failure_Looks_Like" ], "categories": [ "Essays", "Alignment and safety" ] }, "Situational_Awareness_(essay)": { "title": "Situational Awareness (essay)", "type": "essay", "words": 808, "refs": 7, "outbound": [ "AI_2027", "Anthropic", "Chain-of-thought_prompting", "Constitutional_AI", "DeepSeek_R1_release_shock", "Eliciting_Latent_Knowledge", "GPT-2", "GPT-4", "Ilya_Sutskever", "Machines_of_Loving_Grace", "OpenAI", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Seminal_AI_essays", "The_Bitter_Lesson" ], "categories": [ "Essays", "AI governance" ] }, "Sliding-window_attention": { "title": "Sliding-window attention", "type": "technique", "words": 1273, "refs": 7, "outbound": [ "Attention_sink", "FlashAttention", "Gemma", "Hybrid_architecture_(LLM)", "Large_language_model", "Mistral_7B", "Mistral_AI", "OpenAI", "Self-attention", "Softmax", "Sparse_attention", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "Architectures", "Concepts", "Inference" ] }, "SmolLM": { "title": "SmolLM", "type": "model", "words": 359, "refs": 3, "outbound": [ "Direct_preference_optimization", "Gemma", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_3.2", "Mistral_7B", "OLMo", "Pythia", "Qwen2.5", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "2024 model releases", "Open-source AI" ] }, "Softmax": { "title": "Softmax", "type": "concept", "words": 1340, "refs": 6, "outbound": [ "Attention_sink", "Backpropagation", "Cross-entropy_loss", "Knowledge_distillation", "Large_language_model", "Mixtral_8x7B", "Mixture_of_experts", "Multi-head_attention", "Next-token_prediction", "Self-attention", "Sparse_attention", "Transformer_(architecture)" ], "categories": [ "Concepts", "Architectures" ] }, "Software_2.0": { "title": "Software 2.0", "type": "essay", "words": 598, "refs": 3, "outbound": [ "AlphaZero", "Andrej_Karpathy", "Backpropagation", "ImageNet", "Large_language_model", "Pretraining", "Seminal_AI_essays", "The_Unreasonable_Effectiveness_of_Recurrent_Neural_Networks" ], "categories": [ "Essays", "Concepts" ] }, "Sora": { "title": "Sora", "type": "model", "words": 370, "refs": 3, "outbound": [ "DALL-E", "DALL-E_3", "GPT_(model_family)", "Google_DeepMind", "Hailuo", "HunyuanVideo", "Large_language_model", "Meta_AI", "MiniMax", "Movie_Gen", "OpenAI", "Sora_2", "Stable_Diffusion", "Tencent", "Transformer_(architecture)", "Veo" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2024 model releases" ] }, "Sora_2": { "title": "Sora 2", "type": "model", "words": 397, "refs": 3, "outbound": [ "DALL-E_3", "GPT-4o", "Google_DeepMind", "Hailuo", "HunyuanVideo", "Kling", "Meta_AI", "MiniMax", "Movie_Gen", "OpenAI", "Sora", "SynthID", "Tencent", "Veo" ], "categories": [ "OpenAI models", "Models", "Multimodal models", "2025 model releases" ] }, "Sparse_attention": { "title": "Sparse attention", "type": "technique", "words": 1786, "refs": 10, "outbound": [ "Attention_sink", "BERT", "DeepSeek", "DeepSeek-V3", "DeepSeek-V3.1", "DeepSeek_(model_family)", "FlashAttention", "GPT-3", "Gemini_1.5", "Hybrid_architecture_(LLM)", "Large_language_model", "MiniMax-Text-01", "Mistral_7B", "Mixture_of_experts", "Multi-head_attention", "Multi-head_latent_attention", "OpenAI", "PagedAttention", "Self-attention", "Sliding-window_attention", "State-space_model", "Transformer_(architecture)" ], "categories": [ "Architectures", "Concepts", "Inference" ] }, "Sparse_autoencoder": { "title": "Sparse autoencoder", "type": "technique", "words": 2081, "refs": 17, "outbound": [ "Activation_steering", "Activation_verbalizer", "Anthropic", "Chain-of-thought_prompting", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_3", "Claude_Opus_4.6", "Constitutional_AI", "Direct_preference_optimization", "EleutherAI", "Eliciting_Latent_Knowledge", "GPT-2", "GPT-4", "Gemma", "Google_DeepMind", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Llama_(model_family)", "Logit_lens", "Mechanistic_interpretability", "OpenAI", "Pythia", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Self-attention", "Sparse_attention", "Superposition_(interpretability)", "Test-time_compute", "Transformer_(architecture)" ], "categories": [ "Concepts", "Alignment and safety" ] }, "Stability_AI": { "title": "Stability AI", "type": "organization", "words": 619, "refs": 6, "outbound": [ "DALL-E", "EleutherAI", "Google_DeepMind", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Meta_AI", "OpenAI", "Stable_Diffusion", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Stable_Diffusion": { "title": "Stable Diffusion", "type": "model", "words": 413, "refs": 4, "outbound": [ "BLOOM", "CLIP", "DALL-E_2", "Hugging_Face", "HunyuanVideo", "Imagen", "Llama_(model_family)", "Meta_AI", "OpenAI", "Seedream", "Sora", "Stability_AI", "Transformer_(architecture)", "Veo" ], "categories": [ "Models", "2022 model releases", "Open-weight models", "Open-source AI" ] }, "StarCoder": { "title": "StarCoder", "type": "model", "words": 354, "refs": 3, "outbound": [ "Ai2", "BLOOM", "DeepSeek-Coder-V2", "EleutherAI", "Hugging_Face", "Mistral_(model_family)", "NVIDIA", "Qwen_(model_family)", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Models", "Open-weight models", "Code models", "2023 model releases" ] }, "State-space_model": { "title": "State-space model", "type": "concept", "words": 2281, "refs": 22, "outbound": [ "AI21_Labs", "Attention_sink", "Codestral", "EleutherAI", "Falcon_(models)", "FlashAttention", "Gemma", "Google_DeepMind", "Hugging_Face", "Hunyuan", "Hybrid_architecture_(LLM)", "IBM_Granite", "Jamba", "Knowledge_distillation", "Large_language_model", "Llama_2", "MiniMax", "MiniMax-M1", "MiniMax-M2", "MiniMax-Text-01", "Mistral_AI", "Mixtral_8x7B", "Mixture_of_experts", "Moonshot_AI", "Multi-head_attention", "Multi-head_latent_attention", "NVIDIA", "Nemotron", "PagedAttention", "Pythia", "Qwen_team", "RWKV", "Scaling_laws", "Self-attention", "Sliding-window_attention", "Sparse_attention", "Technology_Innovation_Institute", "Tencent", "The_Pile", "Transformer_(architecture)", "Vision_Transformer" ], "categories": [ "Architectures", "Concepts" ] }, "Stated_missions_of_AI_labs": { "title": "Stated missions of AI labs", "type": "reference", "words": 1295, "refs": 8, "outbound": [ "01.AI", "AFM-4.5B", "AI21_Labs", "Ai2", "Aleph_Alpha", "Ant_Group", "Anthropic", "Arcee_AI", "Aya", "BLOOM", "Baichuan", "Baidu", "ByteDance", "Cerebras-GPT", "Cerebras_Systems", "ChatGLM", "Chinese_AI_labs", "Claude_(model_family)", "Claude_Fable_5", "Codestral", "Cohere", "Command_(models)", "Constitutional_AI", "Cursor", "DBRX", "Databricks", "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "Doubao", "ERNIE_5", "ERNIE_Bot", "EleutherAI", "Falcon_(models)", "GLM-4.5", "GLM-5", "GPT-5.6", "GPT-J", "GPT-NeoX-20B", "GPT_(model_family)", "Gemini_(model_family)", "Gemma", "Google_DeepMind", "Grok_(model_family)", "Grok_4.5", "Hailuo", "Huawei", "Hugging_Face", "Hunyuan", "HunyuanVideo", "IDEFICS", "InternLM", "InternVL", "Jamba", "Jurassic_(models)", "Kimi_K2", "Kimi_K3", "LFM_(models)", "Ling_(models)", "Liquid_AI", "Llama_(model_family)", "Llama_4", "LongCat", "Megatron-LM", "Meituan", "Meta_AI", "MiMo", "MiniMax", "MiniMax-M2", "MiniMax-Text-01", "Mistral_(model_family)", "Mistral_AI", "Mixtral_8x7B", "Molmo", "Moondream", "Moondream_(model_family)", "Moonshot_AI", "NVIDIA", "Ndea", "Nemotron", "Nous_Research", "OLMo", "OpenAI", "Prometheus", "Pythia", "QwQ-32B", "Qwen3", "Qwen3-Max", "Qwen_team", "Safe_Superintelligence", "Seed-OSS", "Seedream", "Segment_Anything", "SenseNova", "SenseTime", "Shanghai_AI_Laboratory", "SmolLM", "Sora", "Stability_AI", "Stable_Diffusion", "Step-2", "StepFun", "Technology_Innovation_Institute", "Tencent", "Thinking_Machines_Lab", "Trinity_Large", "Trinity_Large_Thinking", "Tulu_3", "Veo", "Xiaomi", "Yi_(model_family)", "Zhipu_AI", "iFlytek", "iFlytek_Spark", "o1", "xAI" ], "categories": [ "Organizations", "Reference" ] }, "Step-2": { "title": "Step-2", "type": "model", "words": 336, "refs": 3, "outbound": [ "Chinese_AI_labs", "DeepSeek-V2", "Doubao", "ERNIE_Bot", "GLM-4", "GPT-4", "Hunyuan", "Instruction_tuning", "Large_language_model", "MMLU", "MiniMax", "Mixture_of_experts", "Qwen2.5", "StepFun", "Transformer_(architecture)" ], "categories": [ "Models", "2024 model releases", "Mixture-of-experts models" ] }, "StepFun": { "title": "StepFun", "type": "organization", "words": 509, "refs": 5, "outbound": [ "01.AI", "Baichuan", "Chinese_AI_labs", "DeepSeek", "Hugging_Face", "MMLU", "Microsoft_AI", "MiniMax", "Mixture_of_experts", "Moonshot_AI", "Step-2", "Zhipu_AI" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Superposition_(interpretability)": { "title": "Superposition (interpretability)", "type": "concept", "words": 1771, "refs": 19, "outbound": [ "Activation_steering", "Activation_verbalizer", "Anthropic", "CLIP", "Chain-of-thought_prompting", "Circuits_(interpretability)", "Claude_(model_family)", "Claude_3", "Claude_Opus_4.6", "Constitutional_AI", "EleutherAI", "Eliciting_Latent_Knowledge", "GPT-2", "GPT-4", "Gemma", "Google_DeepMind", "Hugging_Face", "Large_language_model", "Logit_lens", "Mechanistic_interpretability", "Mixture_of_experts", "Multi-head_attention", "OpenAI", "Pythia", "Reinforcement_learning_from_human_feedback", "Scaling_laws", "Sparse_autoencoder", "Vision_Transformer" ], "categories": [ "Concepts", "Alignment and safety" ] }, "Supervised_fine-tuning": { "title": "Supervised fine-tuning", "type": "technique", "words": 1349, "refs": 12, "outbound": [ "Ai2", "Alpaca", "BERT", "Chain-of-thought_prompting", "Constitutional_AI", "Cross-entropy_loss", "DeepSeek", "DeepSeek-R1", "DeepSeek-V3", "Direct_preference_optimization", "Flan-T5", "GPT-1", "GPT-3", "Hugging_Face", "InstructGPT", "Instruction_tuning", "Knowledge_distillation", "LLaMA", "Large_language_model", "Llama_(model_family)", "Llama_2", "MMLU", "Meta_AI", "Next-token_prediction", "OpenAI", "Pretraining", "Qwen2.5", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)", "Tulu_3", "Vicuna" ], "categories": [ "Post-training", "Training methods" ] }, "SynthID": { "title": "SynthID", "type": "model", "words": 483, "refs": 3, "outbound": [ "DALL-E_3", "FLUX", "Gemini_(model_family)", "Gemma", "Google_DeepMind", "Hugging_Face", "Imagen", "Kling", "Large_language_model", "Sora", "Veo" ], "categories": [ "Google models", "Models", "2023 model releases", "Multimodal models" ] }, "T5": { "title": "T5", "type": "model", "words": 453, "refs": 4, "outbound": [ "BERT", "Common_Crawl", "Flan-T5", "GPT-2", "GPT-3", "Hugging_Face", "Imagen", "Instruction_tuning", "Large_language_model", "PaLM", "RoBERTa", "Transformer_(architecture)", "XLNet" ], "categories": [ "Google models", "Models", "Open-weight models", "2019 model releases" ] }, "TD-Gammon": { "title": "TD-Gammon", "type": "model", "words": 1281, "refs": 11, "outbound": [ "AlphaGo", "Arthur_Samuel", "Backpropagation", "David_Rumelhart", "Feed-forward_network", "Geoffrey_Hinton", "Perceptron", "Richard_Sutton" ], "categories": [ "Models", "Pre-transformer systems" ] }, "Teacher_forcing": { "title": "Teacher forcing", "type": "concept", "words": 1477, "refs": 6, "outbound": [ "Cross-entropy_loss", "GPT_(model_family)", "LSTM", "Large_language_model", "Next-token_prediction", "Pretraining", "Reinforcement_learning_from_human_feedback", "Seq2seq", "Transformer_(architecture)" ], "categories": [ "Concepts", "Training methods" ] }, "Technology_Innovation_Institute": { "title": "Technology Innovation Institute", "type": "organization", "words": 921, "refs": 7, "outbound": [ "Chinese_AI_labs", "Falcon_(models)", "Hugging_Face", "Hybrid_architecture_(LLM)", "Jamba", "Large_language_model", "Llama_(model_family)", "MMLU", "Meta_AI", "Mistral_AI", "Qwen_(model_family)", "State-space_model", "Stated_missions_of_AI_labs", "The_Pile", "Transformer_(architecture)" ], "categories": [ "Organizations", "Open-source AI" ] }, "Tencent": { "title": "Tencent", "type": "organization", "words": 543, "refs": 5, "outbound": [ "Baidu", "ByteDance", "Chinese_AI_labs", "DeepSeek-R1", "DeepSeek_(model_family)", "Google_DeepMind", "Hugging_Face", "Hunyuan", "HunyuanVideo", "Hunyuan_3.0", "Mixture_of_experts", "Moonshot_AI", "Qwen_(model_family)", "Stability_AI", "Veo" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "Test-time_compute": { "title": "Test-time compute", "type": "concept", "words": 325, "refs": 3, "outbound": [ "Anthropic", "Chain-of-thought_prompting", "DeepSeek", "DeepSeek-R1", "Google_DeepMind", "Kimi_(model_family)", "MMLU", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Scaling_laws", "o1" ], "categories": [ "Concepts", "Reasoning models", "Inference" ] }, "The_Bitter_Lesson": { "title": "The Bitter Lesson", "type": "essay", "words": 876, "refs": 6, "outbound": [ "AI_and_Compute", "AlexNet", "AlphaFold", "AlphaGo", "AlphaZero", "Chain-of-thought_prompting", "GPT-3", "ImageNet", "Large_language_model", "Next-token_prediction", "OpenAI", "Pretraining", "Richard_Sutton", "Scaling_laws", "Seminal_AI_essays", "Situational_Awareness_(essay)", "Test-time_compute", "The_Scaling_Hypothesis", "Transformer_(architecture)", "o1" ], "categories": [ "Essays", "Concepts" ] }, "The_Illustrated_Transformer": { "title": "The Illustrated Transformer", "type": "essay", "words": 504, "refs": 3, "outbound": [ "Cross-entropy_loss", "Embedding_(machine_learning)", "Large_language_model", "Layer_normalization", "Multi-head_attention", "Positional_encoding", "RMSNorm", "Self-attention", "Seminal_AI_essays", "Seq2seq", "Softmax", "Transformer_(architecture)" ], "categories": [ "Essays", "Architectures" ] }, "The_Most_Important_Century": { "title": "The Most Important Century", "type": "essay", "words": 530, "refs": 4, "outbound": [ "AI_2027", "Eliciting_Latent_Knowledge", "Machines_of_Loving_Grace", "No_Fire_Alarm_for_Artificial_General_Intelligence", "Scaling_laws", "Seminal_AI_essays", "Situational_Awareness_(essay)", "What_Failure_Looks_Like" ], "categories": [ "Essays", "AI governance" ] }, "The_Pile": { "title": "The Pile", "type": "dataset", "words": 1146, "refs": 14, "outbound": [ "Cerebras-GPT", "Cerebras_Systems", "Chinchilla", "Comma_(models)", "Common_Crawl", "Common_Pile", "EleutherAI", "GPT-2", "GPT-3", "GPT-J", "GPT-Neo", "GPT-NeoX-20B", "Google_DeepMind", "Hugging_Face", "LLaMA", "Large_language_model", "Llama_2", "Megatron-Turing_NLG", "Meta_AI", "NVIDIA", "OpenAI", "Pythia", "YaLM-100B", "Yandex" ], "categories": [ "Training datasets", "Open-source AI" ] }, "The_Scaling_Hypothesis": { "title": "The Scaling Hypothesis", "type": "essay", "words": 592, "refs": 5, "outbound": [ "AI_2027", "AI_and_Compute", "GPT-2", "GPT-3", "In-context_learning", "OpenAI", "Scaling_laws", "Seminal_AI_essays", "Situational_Awareness_(essay)", "The_Bitter_Lesson" ], "categories": [ "Essays", "Concepts" ] }, "The_Unreasonable_Effectiveness_of_Recurrent_Neural_Networks": { "title": "The Unreasonable Effectiveness of Recurrent Neural Networks", "type": "essay", "words": 633, "refs": 4, "outbound": [ "Andrej_Karpathy", "GPT_(model_family)", "LSTM", "Large_language_model", "Mechanistic_interpretability", "Next-token_prediction", "Seminal_AI_essays", "Simulators_(essay)", "Software_2.0", "Transformer_(architecture)" ], "categories": [ "Essays", "Architectures" ] }, "Thinking_Machines_Lab": { "title": "Thinking Machines Lab", "type": "organization", "words": 1693, "refs": 13, "outbound": [ "Ai2", "Anthropic", "DALL-E", "DeepSeek-V3.1", "Distributed_training", "EleutherAI", "GPT-4o", "Hugging_Face", "Instruction_tuning", "Kimi_(model_family)", "Large_language_model", "Liquid_AI", "Llama_(model_family)", "Meta_AI", "Mistral_AI", "Mixture_of_experts", "Moonshot_AI", "NVIDIA", "Nemotron", "Nous_Research", "OpenAI", "Prometheus", "Qwen3", "Reinforcement_learning_from_human_feedback", "Removal_of_Sam_Altman_from_OpenAI_(annotated_edition)", "Safe_Superintelligence", "Segment_Anything", "Stated_missions_of_AI_labs", "Supervised_fine-tuning", "gpt-oss", "xAI" ], "categories": [ "Organizations", "Frontier labs" ] }, "Tokenization": { "title": "Tokenization", "type": "concept", "words": 1750, "refs": 19, "outbound": [ "BERT", "Embedding_(machine_learning)", "GPT-2", "GPT-3", "GPT-3.5", "GPT-4", "GPT-4o", "Gemma", "Large_language_model", "Llama_(model_family)", "Llama_3", "Next-token_prediction", "OpenAI", "Pretraining", "Qwen_(model_family)", "Softmax", "T5", "Teacher_forcing" ], "categories": [ "Concepts", "Training methods" ] }, "Transformer_(architecture)": { "title": "Transformer (architecture)", "type": "concept", "words": 2743, "refs": 25, "outbound": [ "ALBERT", "Anthropic", "Attention_sink", "BERT", "CLIP", "Chain-of-thought_prompting", "Chinchilla", "Claude_(model_family)", "DALL-E", "DeepSeek-V2", "DeepSeek-V3", "Diffusion_language_model", "Distributed_training", "ELECTRA", "Embedding_(machine_learning)", "Feed-forward_network", "Flan-T5", "FlashAttention", "GLM_(model_family)", "GPT-1", "GPT-2", "GPT-3", "GPT-4", "GPT-5", "GPT_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "Gopher", "Grok_(model_family)", "Hybrid_architecture_(LLM)", "Instruction_tuning", "Kimi_K2", "LaMDA", "Large_language_model", "Layer_normalization", "Llama_3", "MMLU", "Megatron-LM", "Meta_AI", "Mistral_7B", "Mixtral_8x7B", "Mixture_of_experts", "Multi-head_attention", "Multi-head_latent_attention", "NVIDIA", "Next-token_prediction", "OpenAI", "PaLM", "PagedAttention", "Positional_encoding", "Qwen3", "RMSNorm", "Reinforcement_learning_from_human_feedback", "RoBERTa", "Scaling_laws", "Segment_Anything", "Self-attention", "Seq2seq", "Sliding-window_attention", "Softmax", "Sora", "Sparse_attention", "Stable_Diffusion", "State-space_model", "T5", "Test-time_compute", "The_Illustrated_Transformer", "Veo", "Vision_Transformer", "Whisper", "XLNet", "xAI" ], "categories": [ "Architectures", "Concepts" ] }, "Trinity_(model_family)": { "title": "Trinity (model family)", "type": "family", "words": 867, "refs": 7, "outbound": [ "AFM-4.5B", "Arcee_AI", "Claude_Opus_4.6", "DeepSeek-R1", "DeepSeek_(model_family)", "Distributed_training", "GLM-5", "GLM_(model_family)", "Hugging_Face", "Instruction_tuning", "Kimi_(model_family)", "Kimi_K3", "Knowledge_distillation", "Large_language_model", "Llama_(model_family)", "Llama_3.1", "Llama_4", "MMLU", "Meta_AI", "Mixture_of_experts", "NVIDIA", "Scaling_laws", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)", "Trinity_Large", "Trinity_Large_Thinking", "Trinity_Mini", "Trinity_Nano", "o1" ], "categories": [ "Model families", "Open-weight models", "Mixture-of-experts models", "Arcee models" ] }, "Trinity_Large": { "title": "Trinity Large", "type": "model", "words": 854, "refs": 6, "outbound": [ "AFM-4.5B", "Arcee_AI", "Claude_Opus_4.6", "DeepSeek-R1", "Distributed_training", "GLM-5", "Hugging_Face", "Instruction_tuning", "Kimi_K3", "Knowledge_distillation", "Large_language_model", "Llama_3.1", "Llama_4", "MMLU", "Meta_AI", "Mixture_of_experts", "NVIDIA", "Scaling_laws", "Supervised_fine-tuning", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large_Thinking", "Trinity_Mini", "Trinity_Nano" ], "categories": [ "Arcee models", "Models", "2026 model releases", "Open-weight models", "Mixture-of-experts models" ] }, "Trinity_Large_Thinking": { "title": "Trinity Large Thinking", "type": "model", "words": 906, "refs": 6, "outbound": [ "AFM-4.5B", "Arcee_AI", "Claude_Opus_4.6", "DeepSeek-R1", "Distributed_training", "GLM-5", "Hugging_Face", "Instruction_tuning", "Kimi_K3", "Knowledge_distillation", "LLaMA", "Large_language_model", "Llama_3.1", "Llama_4", "MMLU", "Meta_AI", "Mixture_of_experts", "NVIDIA", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Test-time_compute", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Mini", "Trinity_Nano", "o1" ], "categories": [ "Arcee models", "Models", "2026 model releases", "Open-weight models", "Mixture-of-experts models", "Reasoning models" ] }, "Trinity_Mini": { "title": "Trinity Mini", "type": "model", "words": 313, "refs": 3, "outbound": [ "AFM-4.5B", "Arcee_AI", "Hugging_Face", "Large_language_model", "Mixture_of_experts", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Large_Thinking", "Trinity_Nano" ], "categories": [ "Arcee models", "Models", "2025 model releases", "Open-weight models", "Mixture-of-experts models", "Reasoning models" ] }, "Trinity_Nano": { "title": "Trinity Nano", "type": "model", "words": 390, "refs": 4, "outbound": [ "AFM-4.5B", "Arcee_AI", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "Mixture_of_experts", "Transformer_(architecture)", "Trinity_(model_family)", "Trinity_Large", "Trinity_Large_Thinking", "Trinity_Mini" ], "categories": [ "Arcee models", "Models", "2025 model releases", "Mixture-of-experts models", "Open-weight models" ] }, "Tulu_3": { "title": "Tulu 3", "type": "model", "words": 380, "refs": 3, "outbound": [ "Ai2", "DeepSeek-V3", "Direct_preference_optimization", "GPT-4o", "Instruction_tuning", "Llama_3.1", "OLMo", "Qwen2.5", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Transformer_(architecture)" ], "categories": [ "2024 model releases", "Open-weight models", "Open-source AI", "Models" ] }, "Turing-NLG": { "title": "Turing-NLG", "type": "model", "words": 419, "refs": 5, "outbound": [ "GPT-2", "GPT-3", "Large_language_model", "MAI-1", "Megatron-LM", "Megatron-Turing_NLG", "Microsoft_AI", "NVIDIA", "OpenAI", "Scaling_laws", "Transformer_(architecture)" ], "categories": [ "Microsoft models", "Models", "2020 model releases" ] }, "ULMFiT": { "title": "ULMFiT", "type": "model", "words": 326, "refs": 3, "outbound": [ "BERT", "ELMo", "GPT-1", "Instruction_tuning", "LSTM", "Large_language_model", "Pretraining", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Models", "2018 model releases", "Training methods" ] }, "Veo": { "title": "Veo", "type": "model", "words": 368, "refs": 3, "outbound": [ "ByteDance", "Gemini_(model_family)", "Gemini_2.5", "Google_DeepMind", "Hailuo", "HunyuanVideo", "Imagen", "MiniMax", "Movie_Gen", "OpenAI", "Seedream", "Sora", "SynthID", "Tencent" ], "categories": [ "Google models", "Models", "Multimodal models", "2024 model releases" ] }, "Vicuna": { "title": "Vicuna", "type": "model", "words": 363, "refs": 3, "outbound": [ "Alpaca", "GPT-3.5", "GPT-4", "Hugging_Face", "Instruction_tuning", "LLaMA", "Large_language_model", "Llama_(model_family)", "Llama_2", "Meta_AI", "Mistral_7B", "OpenAI", "Transformer_(architecture)" ], "categories": [ "Models", "2023 model releases", "Open-weight models", "Open-source AI" ] }, "Vision_Transformer": { "title": "Vision Transformer", "type": "concept", "words": 2552, "refs": 18, "outbound": [ "AlexNet", "Attention_sink", "BERT", "BLIP", "CLIP", "Claude_(model_family)", "DALL-E_2", "FLUX", "Flamingo", "FlashAttention", "GPT-2", "GPT-3", "GPT-4", "GPT-4o", "Gemini_(model_family)", "Gemini_3", "Gemma", "Google_DeepMind", "Hugging_Face", "IDEFICS", "ImageNet", "InternVL", "Knowledge_distillation", "LLaVA", "Large_language_model", "Llama_3.2", "Meta_AI", "Mistral_AI", "Mixture_of_experts", "Molmo", "Moondream_(model_family)", "Multi-head_attention", "OpenAI", "PaliGemma", "Pixtral", "Qwen2.5-VL", "Scaling_laws", "Segment_Anything", "Self-attention", "Sliding-window_attention", "Softmax", "Sora", "Sparse_attention", "Stable_Diffusion", "Tokenization", "Transformer_(architecture)", "Veo", "Whisper" ], "categories": [ "Architectures", "Multimodal models", "Concepts" ] }, "Walter_Pitts": { "title": "Walter Pitts", "type": "person", "words": 805, "refs": 7, "outbound": [ "Backpropagation", "Feed-forward_network", "Frank_Rosenblatt", "Hopfield_network", "John_Hopfield", "Large_language_model", "Perceptron", "Warren_McCulloch" ], "categories": [ "People", "Computer science pioneers" ] }, "Warren_McCulloch": { "title": "Warren McCulloch", "type": "person", "words": 981, "refs": 11, "outbound": [ "Backpropagation", "Feed-forward_network", "Frank_Rosenblatt", "Hopfield_network", "John_Hopfield", "Large_language_model", "Perceptron", "Walter_Pitts" ], "categories": [ "People", "Computer science pioneers" ] }, "What_Failure_Looks_Like": { "title": "What Failure Looks Like", "type": "essay", "words": 645, "refs": 5, "outbound": [ "AI_2027", "Eliciting_Latent_Knowledge", "Large_language_model", "No_Fire_Alarm_for_Artificial_General_Intelligence", "Reinforcement_learning_from_human_feedback", "Reward_model", "Seminal_AI_essays", "Simulators_(essay)", "Situational_Awareness_(essay)" ], "categories": [ "Essays", "Alignment and safety" ] }, "Whisper": { "title": "Whisper", "type": "model", "words": 374, "refs": 3, "outbound": [ "DALL-E", "GPT-2", "GPT-3", "GPT-3.5", "GPT-4o", "Hugging_Face", "Knowledge_distillation", "Large_language_model", "OpenAI", "Transformer_(architecture)", "gpt-oss" ], "categories": [ "OpenAI models", "Models", "2022 model releases", "Open-weight models", "Open-source AI" ] }, "Word2vec": { "title": "Word2vec", "type": "concept", "words": 377, "refs": 4, "outbound": [ "Anthropic", "BERT", "ELMo", "Embedding_(machine_learning)", "GPT-3", "GloVe", "Large_language_model", "OpenAI", "RoBERTa", "Self-attention", "Transformer_(architecture)" ], "categories": [ "Concepts", "Training methods" ] }, "XLNet": { "title": "XLNet", "type": "model", "words": 387, "refs": 3, "outbound": [ "ALBERT", "BERT", "Common_Crawl", "DistilBERT", "ELECTRA", "GPT-2", "Hugging_Face", "Large_language_model", "RoBERTa", "Self-attention", "T5", "Transformer_(architecture)" ], "categories": [ "Google models", "Models", "2019 model releases", "Open-weight models" ] }, "Xiaomi": { "title": "Xiaomi", "type": "organization", "words": 538, "refs": 5, "outbound": [ "Chinese_AI_labs", "DeepSeek", "DeepSeek_(model_family)", "GLM_(model_family)", "Hugging_Face", "Kimi_(model_family)", "MMLU", "Meituan", "MiMo", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "YaLM-100B": { "title": "YaLM-100B", "type": "model", "words": 515, "refs": 5, "outbound": [ "BLOOM", "Distributed_training", "EleutherAI", "GPT-3", "GPT-NeoX-20B", "Hugging_Face", "Instruction_tuning", "Large_language_model", "Megatron-LM", "Meta_AI", "NVIDIA", "OPT", "The_Pile", "Transformer_(architecture)", "Yandex", "YandexGPT" ], "categories": [ "2022 model releases", "Open-weight models", "Open-source AI", "Models" ] }, "Yandex": { "title": "Yandex", "type": "organization", "words": 1017, "refs": 13, "outbound": [ "Aleph_Alpha", "BLOOM", "Baidu", "GPT_(model_family)", "Hugging_Face", "Large_language_model", "Llama_(model_family)", "Meta_AI", "NVIDIA", "OPT", "Qwen2.5", "Stable_Diffusion", "The_Pile", "Transformer_(architecture)", "YaLM-100B", "YandexGPT" ], "categories": [ "Organizations", "Open-source AI" ] }, "YandexGPT": { "title": "YandexGPT", "type": "model", "words": 668, "refs": 11, "outbound": [ "Chain-of-thought_prompting", "Direct_preference_optimization", "GPT-4o", "Hugging_Face", "Large_language_model", "Qwen2.5", "Reinforcement_learning_from_human_feedback", "Supervised_fine-tuning", "Transformer_(architecture)", "YaLM-100B", "Yandex" ], "categories": [ "Models", "2023 model releases" ] }, "Yann_LeCun": { "title": "Yann LeCun", "type": "person", "words": 2332, "refs": 22, "outbound": [ "AlexNet", "Backpropagation", "Boltzmann_machine", "David_Rumelhart", "Demis_Hassabis", "Frank_Rosenblatt", "GPT_(model_family)", "Geoffrey_Hinton", "Hopfield_network", "ImageNet", "John_Hopfield", "Large_language_model", "Llama_(model_family)", "Meta_AI", "Next-token_prediction", "OpenAI", "Paul_Werbos", "Perceptron", "Reinforcement_learning_from_human_feedback", "Segment_Anything", "Supervised_fine-tuning", "Transformer_(architecture)", "Vision_Transformer", "Yoshua_Bengio" ], "categories": [ "People", "Machine learning researchers" ] }, "Yi_(model_family)": { "title": "Yi (model family)", "type": "model", "words": 351, "refs": 3, "outbound": [ "01.AI", "Baichuan", "ChatGLM", "DeepSeek_LLM", "Hugging_Face", "Instruction_tuning", "InternLM", "LLaMA", "Large_language_model", "MMLU", "Mistral_7B", "Mixtral_8x7B", "Qwen_(model_family)", "Qwen_1", "Transformer_(architecture)" ], "categories": [ "01.AI models", "Models", "Open-weight models", "2023 model releases", "Chinese AI labs" ] }, "Yoshua_Bengio": { "title": "Yoshua Bengio", "type": "person", "words": 2362, "refs": 23, "outbound": [ "AlexNet", "BERT", "Backpropagation", "Boltzmann_machine", "Demis_Hassabis", "Embedding_(machine_learning)", "GPT-4", "GPT_(model_family)", "Geoffrey_Hinton", "GloVe", "Ian_Goodfellow", "Ilya_Sutskever", "ImageNet", "John_Hopfield", "LSTM", "Large_language_model", "Next-token_prediction", "OpenAI", "Pretraining", "Self-attention", "Seq2seq", "Teacher_forcing", "Transformer_(architecture)", "Word2vec", "Yann_LeCun" ], "categories": [ "People", "Machine learning researchers" ] }, "Zhipu_AI": { "title": "Zhipu AI", "type": "organization", "words": 597, "refs": 6, "outbound": [ "ChatGLM", "DeepSeek", "DeepSeek_(model_family)", "GLM-130B", "GLM-4.5", "GLM-4.6", "GLM-5", "GLM_(model_family)", "Hugging_Face", "Large_language_model", "MMLU", "Mixture_of_experts", "Moonshot_AI", "Qwen_(model_family)", "Reinforcement_learning_from_human_feedback", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "gpt-oss": { "title": "gpt-oss", "type": "model", "words": 355, "refs": 4, "outbound": [ "Chain-of-thought_prompting", "Chinese_AI_labs", "DeepSeek-R1", "GPT-2", "GPT-5", "GPT_(model_family)", "Hugging_Face", "Kimi_K2", "Mixture_of_experts", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Supervised_fine-tuning", "Transformer_(architecture)", "o4-mini" ], "categories": [ "OpenAI models", "Models", "Open-weight models", "Mixture-of-experts models", "Reasoning models", "2025 model releases" ] }, "iFlytek": { "title": "iFlytek", "type": "organization", "words": 511, "refs": 4, "outbound": [ "Chinese_AI_labs", "DeepSeek_(model_family)", "GPT-4", "Huawei", "LongCat", "MMLU", "Meituan", "NVIDIA", "Qwen_(model_family)", "SenseTime", "iFlytek_Spark" ], "categories": [ "Organizations", "Chinese AI labs" ] }, "iFlytek_Spark": { "title": "iFlytek Spark", "type": "model", "words": 349, "refs": 4, "outbound": [ "Baidu", "ChatGLM", "DeepSeek_LLM", "Doubao", "ERNIE_Bot", "GPT-3.5", "GPT-4", "Huawei", "Hunyuan", "Instruction_tuning", "Kimi_K2", "Large_language_model", "NVIDIA", "OpenAI", "Qwen2", "SenseNova", "SenseTime", "Transformer_(architecture)", "Yi_(model_family)", "Zhipu_AI", "iFlytek", "o3" ], "categories": [ "iFlytek models", "Models", "2023 model releases", "Chinese AI labs", "Multimodal models" ] }, "o1": { "title": "o1", "type": "model", "words": 303, "refs": 3, "outbound": [ "Anthropic", "Chain-of-thought_prompting", "Claude_(model_family)", "DeepSeek", "DeepSeek-R1", "GLM_(model_family)", "GPT-4", "GPT-5", "GPT_(model_family)", "Gemini_(model_family)", "Google_DeepMind", "Kimi_(model_family)", "OpenAI", "Qwen_(model_family)", "Reinforcement_learning_with_verifiable_rewards", "Test-time_compute", "Transformer_(architecture)", "o3" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2024 model releases" ] }, "o3": { "title": "o3", "type": "model", "words": 371, "refs": 4, "outbound": [ "Chain-of-thought_prompting", "DeepSeek-R1", "GPT-4o", "GPT-5", "GPT-5.1", "GPT_(model_family)", "Gemini_2.5", "Grok_3", "OpenAI", "Test-time_compute", "Transformer_(architecture)", "gpt-oss", "o1", "o4-mini" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2024 model releases", "2025 model releases" ] }, "o4-mini": { "title": "o4-mini", "type": "model", "words": 373, "refs": 3, "outbound": [ "Chain-of-thought_prompting", "GPT-4o", "GPT-4o_mini", "GPT-5", "GPT_(model_family)", "Large_language_model", "OpenAI", "Test-time_compute", "Transformer_(architecture)", "gpt-oss", "o1", "o3" ], "categories": [ "OpenAI models", "Models", "Reasoning models", "2025 model releases" ] }, "xAI": { "title": "xAI", "type": "organization", "words": 569, "refs": 6, "outbound": [ "Anthropic", "DeepSeek", "Google_DeepMind", "Grok-1", "Grok_(model_family)", "Grok_3", "Grok_4", "Hugging_Face", "Large_language_model", "MMLU", "Meta_AI", "Mixture_of_experts", "OpenAI", "Reinforcement_learning_from_human_feedback", "Transformer_(architecture)" ], "categories": [ "Organizations", "Frontier labs" ] } }, "redirects": {} }