# 参考书目与文献索引 (References and Bibliography)

Chapter from the Chinese-language book Anatomy of the Stochastic Parrot: 参考书目与文献索引 (References and Bibliography).

## Metadata

- HTML: https://glenzli.com/en/dr-stochastic-parrot/stochastic-parrot-anatomy/vol-01/source-notes/
- Markdown: https://glenzli.com/en/dr-stochastic-parrot/stochastic-parrot-anatomy/vol-01/source-notes.md
- Collection: Dr. Stochastic Parrot
- Language: en
- Published: 2026-07-17
- Status: published
- Tags: ai-generated, stochastic-parrot, stochastic-parrot-anatomy, ai

## Content

本列表收录书中引用的经典论文、核心书籍、技术报告与协议规范，按主题分类整理。2024 年以后仍可能变化的模型、产品和协议条目校准至 **2026 年 7 月 12 日**；动态文档的日期表示本书核验口径，而不等同于论文发表时间。

## 1. 基础理论 (Foundations)

*   **[McCulloch & Pitts, 1943]** *A Logical Calculus of the Ideas Immanent in Nervous Activity*. (M-P 神经元模型的提出)

*   **[Turing, 1950]** *Computing Machinery and Intelligence*. (图灵测试)

*   **[McCarthy et al., 1955]** [*A Proposal for the Dartmouth Summer Research Project on Artificial Intelligence*](http://www-formal.stanford.edu/jmc/history/dartmouth/dartmouth.html). (人工智能夏季研究项目提案)

*   **[Rosenblatt, 1958]** *The Perceptron: A Probabilistic Model for Information Storage and Organization in the Brain*. (感知机)

*   **[Novikoff, 1962]** [*On Convergence Proofs for Perceptrons*](https://cs.uwaterloo.ca/~y328yu/classics/novikoff.pdf). (感知机错误次数界的经典证明)

*   **[Minsky & Papert, 1969]** *Perceptrons*. (单层感知机能力边界及相关分析)

*   **[Rumelhart, Hinton & Williams, 1986]** *Learning representations by back-propagating errors*. (反向传播的经典应用)

*   **[Cybenko, 1989]** [*Approximation by Superpositions of a Sigmoidal Function*](https://doi.org/10.1007/BF02551274). (连续 sigmoidal/discriminatory 激活的一致逼近定理)

*   **[Hornik, 1991]** [*Approximation Capabilities of Multilayer Feedforward Networks*](https://doi.org/10.1016/0893-6080(91)90009-T). (前馈网络逼近能力的推广)

*   **[Leshno et al., 1993]** [*Multilayer Feedforward Networks with a Nonpolynomial Activation Function Can Approximate Any Function*](https://doi.org/10.1016/S0893-6080(05)80131-5). (非多项式激活条件的刻画)

*   **[Cortes & Vapnik, 1995]** [*Support-Vector Networks*](https://doi.org/10.1007/BF00994018). (软间隔支持向量机)

*   **[Breiman, 2001]** [*Random Forests*](https://doi.org/10.1023/A:1010933404324). (随机森林)

*   **[Friedman, 2001]** [*Greedy Function Approximation: A Gradient Boosting Machine*](https://doi.org/10.1214/aos/1013203451). (梯度提升)

### 1.1 统计学习与随机逼近

*   **[Robbins & Monro, 1951]** [*A Stochastic Approximation Method*](https://doi.org/10.1214/aoms/1177729586). (随机逼近与衰减步长框架)

*   **[Hoeffding, 1963]** [*Probability Inequalities for Sums of Bounded Random Variables*](https://doi.org/10.1080/01621459.1963.10500830). (有界独立随机变量的集中不等式)

*   **[Vapnik & Chervonenkis, 1971]** [*On the Uniform Convergence of Relative Frequencies of Events to Their Probabilities*](https://doi.org/10.1137/1116025). (VC 维与一致收敛)

*   **[Sauer, 1972]** [*On the Density of Families of Sets*](https://doi.org/10.1016/0097-3165(72)90019-2). (增长函数的组合上界)

### 1.2 优化、初始化与正则化

*   **[Glorot & Bengio, 2010]** [*Understanding the difficulty of training deep feedforward neural networks*](https://proceedings.mlr.press/v9/glorot10a.html). (Xavier 初始化与信号尺度)

*   **[He et al., 2015]** [*Delving Deep into Rectifiers*](https://arxiv.org/abs/1502.01852). (适配 rectifier 的初始化)

*   **[Srivastava et al., 2014]** [*Dropout: A Simple Way to Prevent Neural Networks from Overfitting*](https://jmlr.org/papers/v15/srivastava14a.html). (dropout)

*   **[Kingma & Ba, 2014]** [*Adam: A Method for Stochastic Optimization*](https://arxiv.org/abs/1412.6980). (Adam)

*   **[Loshchilov & Hutter, 2019]** [*Decoupled Weight Decay Regularization*](https://arxiv.org/abs/1711.05101). (AdamW 与解耦 weight decay)

## 2. 卷积神经网络 (CNNs)

*   **[LeCun et al., 1998]** *Gradient-based learning applied to document recognition*. (LeNet-5)

*   **[Krizhevsky et al., 2012]** *ImageNet Classification with Deep Convolutional Neural Networks*. (AlexNet)
*   **[Simonyan & Zisserman, 2014]** *Very Deep Convolutional Networks for Large-Scale Image Recognition*. (VGGNet)

*   **[He et al., 2016]** *Deep Residual Learning for Image Recognition*. (ResNet: 残差连接)

## 3. 序列模型与 RNN (Sequence Models)

*   **[Hochreiter & Schmidhuber, 1997]** *Long Short-Term Memory*. (LSTM 的提出)

*   **[Cho et al., 2014]** *Learning Phrase Representations using RNN Encoder-Decoder for Statistical Machine Translation*. (GRU 与 Seq2Seq)
*   **[Bahdanau, Cho & Bengio, 2014]** [*Neural Machine Translation by Jointly Learning to Align and Translate*](https://arxiv.org/abs/1409.0473). (加性注意力)
*   **[Luong, Pham & Manning, 2015]** [*Effective Approaches to Attention-based Neural Machine Translation*](https://arxiv.org/abs/1508.04025). (乘性注意力)

*   **[Peters et al., 2018]** [*Deep Contextualized Word Representations*](https://arxiv.org/abs/1802.05365). (ELMo: 动态词向量)

## 4. Transformer 与预训练语言模型 (Transformer & PLMs)

*   **[Vaswani et al., 2017]** [*Attention Is All You Need*](https://arxiv.org/abs/1706.03762). (Transformer 架构)
*   **[Ba, Kiros & Hinton, 2016]** [*Layer Normalization*](https://arxiv.org/abs/1607.06450). (LayerNorm)

*   **[Zhang & Sennrich, 2019]** [*Root Mean Square Layer Normalization*](https://arxiv.org/abs/1910.07467). (RMSNorm)

*   **[Shazeer, 2020]** [*GLU Variants Improve Transformer*](https://arxiv.org/abs/2002.05202). (GLU、GEGLU 与 SwiGLU 变体)

*   **[Shazeer, 2019]** [*Fast Transformer Decoding: One Write-Head is All You Need*](https://arxiv.org/abs/1911.02150). (Multi-Query Attention)

*   **[Ainslie et al., 2023]** [*GQA: Training Generalized Multi-Query Transformer Models from Multi-Head Checkpoints*](https://arxiv.org/abs/2305.13245). (Grouped-Query Attention)

*   **[Su et al., 2021]** [*RoFormer: Enhanced Transformer with Rotary Position Embedding*](https://arxiv.org/abs/2104.09864). (RoPE)
*   **[Chen et al., 2023]** [*Extending Context Window of Large Language Models via Positional Interpolation*](https://arxiv.org/abs/2306.15595). (位置插值与上下文扩展)
*   **[Liu et al., 2023]** [*Ring Attention with Blockwise Transformers for Near-Infinite Context*](https://arxiv.org/abs/2310.01889). (分布式长上下文注意力)
*   **[Child et al., 2019]** [*Generating Long Sequences with Sparse Transformers*](https://arxiv.org/abs/1904.10509). (稀疏注意力与长序列生成)

*   **[Dao et al., 2022]** [*FlashAttention: Fast and Memory-Efficient Exact Attention with IO-Awareness*](https://arxiv.org/abs/2205.14135). (IO-aware 精确注意力)
*   **[Dao, 2023]** [*FlashAttention-2: Faster Attention with Better Parallelism and Work Partitioning*](https://arxiv.org/abs/2307.08691). (更高并行度的注意力 kernel)

*   **[Fedus, Zoph & Shazeer, 2021]** [*Switch Transformers: Scaling to Trillion Parameter Models with Simple and Efficient Sparsity*](https://arxiv.org/abs/2101.03961). (稀疏专家、容量与路由均衡)

*   **[Gu & Dao, 2023]** [*Mamba: Linear-Time Sequence Modeling with Selective State Spaces*](https://arxiv.org/abs/2312.00752). (选择性状态空间模型)

*   **[Dao & Gu, 2024]** [*Transformers are SSMs: Generalized Models and Efficient Algorithms Through Structured State Space Duality*](https://arxiv.org/abs/2405.21060). (Mamba-2 与 SSM/Attention 统一视角)

*   **[Devlin et al., 2018]** [*BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding*](https://arxiv.org/abs/1810.04805). (BERT)

*   **[Liu et al., 2019]** [*RoBERTa: A Robustly Optimized BERT Pretraining Approach*](https://arxiv.org/abs/1907.11692). (移除 NSP 并联合调整数据与训练配方)

*   **[Radford et al., 2018]** [*Improving Language Understanding by Generative Pre-Training*](https://cdn.openai.com/research-covers/language-unsupervised/language_understanding_paper.pdf). (GPT-1)

*   **[Radford et al., 2019]** [*Language Models are Unsupervised Multitask Learners*](https://cdn.openai.com/better-language-models/language_models_are_unsupervised_multitask_learners.pdf). (GPT-2)

*   **[Brown et al., 2020]** [*Language Models are Few-Shot Learners*](https://arxiv.org/abs/2005.14165). (GPT-3)

*   **[Raffel et al., 2020]** [*Exploring the Limits of Transfer Learning with a Unified Text-to-Text Transformer*](https://arxiv.org/abs/1910.10683). (T5)

*   **[Lewis et al., 2019]** [*BART: Denoising Sequence-to-Sequence Pre-training for Natural Language Generation, Translation, and Comprehension*](https://arxiv.org/abs/1910.13461). (BART)

*   **[Kudo & Richardson, 2018]** [*SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processing*](https://aclanthology.org/D18-2012/). (从原始句子训练的子词 tokenizer)

*   **[Loshchilov & Hutter, 2019]** [*Decoupled Weight Decay Regularization*](https://arxiv.org/abs/1711.05101). (AdamW 与解耦 weight decay)

*   **[Lee et al., 2022]** [*Deduplicating Training Data Makes Language Models Better*](https://aclanthology.org/2022.acl-long.577/). (训练语料与 train-test 重复)

## 5. 大模型对齐与优化 (Alignment & Optimization)

*   **[Wei et al., 2021]** *Finetuned Language Models Are Zero-Shot Learners*. (FLAN: 指令微调)

*   **[Ouyang et al., 2022]** [*Training language models to follow instructions with human feedback*](https://arxiv.org/abs/2203.02155). (InstructGPT: RLHF 的应用)

*   **[Schulman et al., 2017]** [*Proximal Policy Optimization Algorithms*](https://arxiv.org/abs/1707.06347). (PPO 算法)
*   **[Amodei et al., 2016]** [*Concrete Problems in AI Safety*](https://arxiv.org/abs/1606.06565). (reward hacking、负副作用与安全问题)

*   **[Rafailov et al., 2023]** [*Direct Preference Optimization: Your Language Model is Secretly a Reward Model*](https://arxiv.org/abs/2305.18290). (DPO)
*   **[Li & Liang, 2021]** [*Prefix-Tuning: Optimizing Continuous Prompts for Generation*](https://arxiv.org/abs/2101.00190). (每层 prefix 参数)
*   **[Lester, Al-Rfou & Constant, 2021]** [*The Power of Scale for Parameter-Efficient Prompt Tuning*](https://arxiv.org/abs/2104.08691). (输入层 soft prompt)

*   **[Hu et al., 2021]** [*LoRA: Low-Rank Adaptation of Large Language Models*](https://arxiv.org/abs/2106.09685). (LoRA)

*   **[Dettmers et al., 2023]** [*QLoRA: Efficient Finetuning of Quantized LLMs*](https://arxiv.org/abs/2305.14314). (QLoRA)
*   **[Frantar et al., 2022]** [*GPTQ: Accurate Post-Training Quantization for Generative Pre-trained Transformers*](https://arxiv.org/abs/2210.17323). (GPTQ 权重量化)
*   **[Xiao et al., 2022]** [*SmoothQuant: Accurate and Efficient Post-Training Quantization for Large Language Models*](https://arxiv.org/abs/2211.10438). (激活/权重量化平滑)
*   **[Lin et al., 2023]** [*AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration*](https://arxiv.org/abs/2306.00978). (AWQ 混合/保护式量化)

*   **[Kaplan et al., 2020]** [*Scaling Laws for Neural Language Models*](https://arxiv.org/abs/2001.08361). (缩放定律)

*   **[Hoffmann et al., 2022]** [*Training Compute-Optimal Large Language Models*](https://arxiv.org/abs/2203.15556). (固定计算预算下的模型规模与训练 token 配置)

*   **[Kwon et al., 2023]** [*Efficient Memory Management for Large Language Model Serving with PagedAttention*](https://arxiv.org/abs/2309.06180). (PagedAttention 与 vLLM)
*   **[DeepSeek-AI, 2024]** [*DeepSeek-V3 Technical Report*](https://arxiv.org/abs/2412.19437). (MLA、MoE 与高效训练)
*   **[DeepSeek-AI, 2025]** [*Native Sparse Attention: Hardware-Aligned and Natively Trainable Sparse Attention*](https://arxiv.org/abs/2502.11089). (NSA: 硬件对齐、可训练稀疏注意力)
*   **[DeepSeek-AI, 2025]** [*DeepSeek-V3.2-Exp: Boosting Long-Context Efficiency with DeepSeek Sparse Attention*](https://huggingface.co/deepseek-ai/DeepSeek-V3.2-Exp). (DSA 实验性长上下文效率模型)
*   **[Zhou et al., 2026]** [*MISA: Mixture of Indexer Sparse Attention for Long-Context LLM Inference*](https://arxiv.org/abs/2605.07363). (独立后续研究：DSA 索引器的 head 轴路由加速，并非 DeepSeek 团队论文)

*   **[Shao et al., 2024]** [*DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models*](https://arxiv.org/abs/2402.03300). (GRPO 的提出)

*   **[DeepSeek-AI, 2025]** [*DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning*](https://arxiv.org/abs/2501.12948). (推理模型、GRPO 与 RL 后训练)
*   **[DeepSeek-AI, 2025]** [*DeepSeek-OCR: Contexts Optical Compression*](https://arxiv.org/abs/2510.18234). (作者所称的 initial investigation：视觉 token 压缩与 OCR 重建，不代表通用长上下文压缩已解决)
*   **[Cheng et al., 2026]** [*Conditional Memory via Scalable Lookup: A New Axis of Sparsity for Large Language Models*](https://arxiv.org/abs/2601.07372). (Engram：哈希寻址 N-gram embedding 与层内条件记忆)

*   **[OpenAI, 2024]** [*Learning to Reason with LLMs*](https://openai.com/index/learning-to-reason-with-llms/). (o1、强化学习与测试时计算的公开说明)
*   **[Dubey et al., 2024]** [*The Llama 3 Herd of Models*](https://arxiv.org/abs/2407.21783). (开放权重模型、预训练与后训练技术报告)
*   **[OpenAI, 2026-07-09]** [*Introducing GPT-5.6*](https://openai.com/index/gpt-5-6/). (GPT-5.6 Sol / Terra / Luna 已 GA；图像输入等能力以发布页为准；核验于 2026-07-12)
*   **[DeepSeek, 2026]** [*DeepSeek API Change Log*](https://api-docs.deepseek.com/updates) 与 [*Models and Pricing*](https://api-docs.deepseek.com/quick_start/pricing). (`deepseek-v4-pro` / `deepseek-v4-flash` 官方 API 标识；核验于 2026-07-12)
*   **[Hinton et al., 2015]** [*Distilling the Knowledge in a Neural Network*](https://arxiv.org/abs/1503.02531). (知识蒸馏)
*   **[Wang et al., 2022]** [*Self-Instruct: Aligning Language Models with Self-Generated Instructions*](https://arxiv.org/abs/2212.10560). (合成指令数据)

*   **[Maini et al., 2024]** [*TOFU: A Task of Fictitious Unlearning for LLMs*](https://openreview.net/forum?id=q0eyIBnE2t). (面向个体虚构资料的遗忘集/保留集、多指标评测与重训参照；基准结果不证明任意观察下的精确遗忘)

*   **[Guo et al., 2020]** [*Certified Data Removal from Machine Learning Models*](https://proceedings.mlr.press/v119/guo20c.html). (以重训输出分布为参照的 certified removal 定义及线性分类器机制)

*   **[Mitchell et al., 2019]** [*Model Cards for Model Reporting*](https://arxiv.org/abs/1810.03993). (模型用途、评测条件、群体差异与限制的结构化发布文档)

*   **[Gebru et al., 2021]** [*Datasheets for Datasets*](https://arxiv.org/abs/1803.09010). (数据动机、组成、收集、处理、用途和维护的结构化文档框架)
*   **[Bai et al., 2022]** [*Constitutional AI: Harmlessness from AI Feedback*](https://arxiv.org/abs/2212.08073). (AI 反馈与安全后训练)
*   **[Korbak et al., 2025]** [*Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety*](https://arxiv.org/abs/2507.11473). (CoT 监控与安全训练风险)
*   **[Wortsman et al., 2022]** [*Model soups: averaging weights of multiple fine-tuned models improves accuracy without increasing inference time*](https://arxiv.org/abs/2203.05482). (权重平均与模型合并)
*   **[Ilharco et al., 2022]** [*Editing Models with Task Arithmetic*](https://arxiv.org/abs/2212.04089). (任务向量与模型编辑)
*   **[Yadav et al., 2023]** [*TIES-Merging: Resolving Interference When Merging Models*](https://arxiv.org/abs/2306.01708). (模型合并干扰处理)
*   **[Yu et al., 2023]** [*Language Models are Super Mario: Absorbing Abilities from Homologous Models as a Free Lunch*](https://arxiv.org/abs/2311.03099). (DARE 与模型合并)

*   **[Leviathan et al., 2023]** [*Fast Inference from Transformers via Speculative Decoding*](https://arxiv.org/abs/2211.17192). (保持目标分布的投机解码)
*   **[Cai et al., 2024]** [*Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads*](https://arxiv.org/abs/2401.10774). (多解码头推理加速)
*   **[Li et al., 2024]** [*EAGLE: Speculative Sampling Requires Rethinking Feature Uncertainty*](https://arxiv.org/abs/2401.15077). (特征级投机解码)

## 6. 多模态与智能体 (Multimodal & Agents)

*   **[Dosovitskiy et al., 2020]** [*An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale*](https://arxiv.org/abs/2010.11929). (ViT)

*   **[Radford et al., 2021]** [*Learning Transferable Visual Models From Natural Language Supervision*](https://arxiv.org/abs/2103.00020). (CLIP)

*   **[van den Oord, Li & Vinyals, 2018]** [*Representation Learning with Contrastive Predictive Coding*](https://arxiv.org/abs/1807.03748). (InfoNCE 与对比预测编码)

*   **[Liu et al., 2023]** [*Visual Instruction Tuning*](https://arxiv.org/abs/2304.08485). (LLaVA)

*   **[Alayrac et al., 2022]** [*Flamingo: a Visual Language Model for Few-Shot Learning*](https://arxiv.org/abs/2204.14198). (冻结视觉与语言组件之间的 Perceiver Resampler、门控交叉注意力及交错图文训练)

*   **[Li et al., 2023]** [*BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Models*](https://arxiv.org/abs/2301.12597). (Q-Former 作为冻结图像编码器与语言模型之间的查询瓶颈)

*   **[Radford et al., 2022]** [*Robust Speech Recognition via Large-Scale Weak Supervision*](https://arxiv.org/abs/2212.04356). (Whisper 的 log-Mel 输入、编码器—解码器结构与多任务语音训练)

*   **[Kingma & Welling, 2013]** [*Auto-Encoding Variational Bayes*](https://arxiv.org/abs/1312.6114). (证据下界、重参数化估计与变分自编码器)

*   **[Goodfellow et al., 2014]** [*Generative Adversarial Nets*](https://arxiv.org/abs/1406.2661). (原始 GAN 极小极大目标、最优判别器与 Jensen--Shannon 散度分析)

*   **[Ho, Jain & Abbeel, 2020]** [*Denoising Diffusion Probabilistic Models*](https://arxiv.org/abs/2006.11239). (离散高斯前向过程、反向模型与简化噪声预测目标)

*   **[Song, Meng & Ermon, 2020]** [*Denoising Diffusion Implicit Models*](https://arxiv.org/abs/2010.02502). (与 DDPM 共享训练目标的非 Markov 采样过程及确定性极限)

*   **[Song et al., 2021]** [*Score-Based Generative Modeling through Stochastic Differential Equations*](https://arxiv.org/abs/2011.13456). (反向时间 SDE、概率流 ODE 与连续时间 score 模型)

*   **[Austin et al., 2021]** [*Structured Denoising Diffusion Models in Discrete State-Spaces*](https://arxiv.org/abs/2107.03006). (D3PM 的离散转移矩阵与反向去噪)

*   **[Rombach et al., 2022]** [*High-Resolution Image Synthesis with Latent Diffusion Models*](https://arxiv.org/abs/2112.10752). (Latent Diffusion / Stable Diffusion 基础)

*   **[Peebles & Xie, 2023]** [*Scalable Diffusion Models with Transformers*](https://arxiv.org/abs/2212.09748). (DiT)

*   **[Lipman et al., 2022]** [*Flow Matching for Generative Modeling*](https://arxiv.org/abs/2210.02747). (Flow Matching 与连续生成路径)

*   **[Heusel et al., 2017]** [*GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium*](https://arxiv.org/abs/1706.08500). (Fréchet Inception Distance 的提出及其特征高斯矩定义)

*   **[Bińkowski et al., 2018]** [*Demystifying MMD GANs*](https://arxiv.org/abs/1801.01401). (Kernel Inception Distance、无偏 MMD 估计及有限样本比较)
*   **[Betker et al., 2023]** [*Improving Image Generation with Better Captions*](https://cdn.openai.com/papers/dall-e-3.pdf). (DALL-E 3 与高质量图文描述)

*   **[OpenAI, 2024]** [*GPT-4o System Card*](https://cdn.openai.com/gpt-4o-system-card.pdf). (文本、视觉、音频的端到端多模态评测与安全分析)

*   **[OpenAI, 2024]** [*Video Generation Models as World Simulators*](https://openai.com/index/video-generation-models-as-world-simulators/). (Sora：时空 patch 上的扩散 Transformer；技术细节披露有限)
*   **[Google DeepMind, 核验于 2026-07-12]** [*Veo*](https://deepmind.google/models/veo/). (视频生成模型族的官方能力材料；不据此推断未公开架构)

*   **[Kondratyuk et al., 2023]** [*VideoPoet: A Large Language Model for Zero-Shot Video Generation*](https://arxiv.org/abs/2312.14125). (视频生成的语言模型路线)

*   **[Reid et al., 2024]** [*Gemini 1.5: Unlocking Multimodal Understanding Across Millions of Tokens of Context*](https://arxiv.org/abs/2403.05530). (长上下文多模态理解)

*   **[Google, 核验于 2026-07-12]** [*Gemini 3.5 Flash Model Documentation*](https://ai.google.dev/gemini-api/docs/models/gemini-3.5-flash). (stable/GA；text/image/video/audio/PDF 输入、text 输出；不支持 Live API)

*   **[Ha & Schmidhuber, 2018]** [*World Models*](https://arxiv.org/abs/1803.10122). (世界模型经典工作)

*   **[Hafner et al., 2019]** [*Dream to Control: Learning Behaviors by Latent Imagination*](https://arxiv.org/abs/1912.01603). (Dreamer 与潜在想象)

*   **[Hafner et al., 2023]** [*Mastering Diverse Domains through World Models*](https://arxiv.org/abs/2301.04104). (DreamerV3 的 RSSM、想象 actor--critic 与跨域训练配方)

*   **[Hansen et al., 2024]** [*TD-MPC2: Scalable, Robust World Models for Continuous Control*](https://arxiv.org/abs/2310.16828). (潜空间局部模型、价值学习与基于 CEM 的模型预测控制)

*   **[Assran et al., 2023]** [*Self-Supervised Learning from Images with a Joint-Embedding Predictive Architecture*](https://arxiv.org/abs/2301.08243). (I-JEPA)

*   **[Bruce et al., 2024]** [*Genie: Generative Interactive Environments*](https://arxiv.org/abs/2402.15391). (从视频学习可交互环境)

*   **[Zitkovich et al., 2023]** [*RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control*](https://proceedings.mlr.press/v229/zitkovich23a.html). (把视觉、语言和机器人动作接入同一 VLA 接口；具体实验不外推为开放环境安全保证)

*   **[Kim et al., 2024]** [*OpenVLA: An Open-Source Vision-Language-Action Model*](https://arxiv.org/abs/2406.09246). (开放 VLA 模型、动作 token 化与跨机器人数据训练)

*   **[Chi et al., 2023]** [*Diffusion Policy: Visuomotor Policy Learning via Action Diffusion*](https://arxiv.org/abs/2303.04137). (用条件 diffusion 生成连续动作序列与 receding-horizon 执行)

*   **[Google DeepMind et al., 2025]** [*Gemini Robotics: Bringing AI into the Physical World*](https://arxiv.org/abs/2503.20020). (VLA 通用模型、机器人形态适配与物理行动研究入口)

*   **[Wei et al., 2022]** [*Chain-of-Thought Prompting Elicits Reasoning in Large Language Models*](https://arxiv.org/abs/2201.11903). (CoT)

*   **[Kojima et al., 2022]** [*Large Language Models are Zero-Shot Reasoners*](https://arxiv.org/abs/2205.11916). (Zero-shot CoT)

*   **[Yao et al., 2022]** [*ReAct: Synergizing Reasoning and Acting in Language Models*](https://arxiv.org/abs/2210.03629). (ReAct 框架)

*   **[Schick et al., 2023]** [*Toolformer: Language Models Can Teach Themselves to Use Tools*](https://arxiv.org/abs/2302.04761). (在语言建模训练中学习 API 调用位置、工具与参数)

*   **[Zhou et al., 2024]** [*WebArena: A Realistic Web Environment for Building Autonomous Agents*](https://arxiv.org/abs/2307.13854). (可重置网站环境、长程任务和功能正确性评测；结果受其网站与任务快照约束)

*   **[Greshake et al., 2023]** [*Not What You've Signed Up For: Compromising Real-World LLM-Integrated Applications with Indirect Prompt Injection*](https://arxiv.org/abs/2302.12173). (攻击者控制的外部数据通过有权限应用触发间接 prompt injection)

*   **[Lewis et al., 2020]** [*Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks*](https://arxiv.org/abs/2005.11401). (RAG)

*   **[Robertson & Zaragoza, 2009]** [*The Probabilistic Relevance Framework: BM25 and Beyond*](https://www.staff.city.ac.uk/~sbrp622/papers/foundations_bm25_review.pdf). (BM25 的概率相关性框架、词频饱和与文档长度归一化)

*   **[Karpukhin et al., 2020]** [*Dense Passage Retrieval for Open-Domain Question Answering*](https://arxiv.org/abs/2004.04906). (DPR 双编码器、困难负例与最大内积检索)

*   **[Khattab & Zaharia, 2020]** [*ColBERT: Efficient and Effective Passage Search via Contextualized Late Interaction over BERT*](https://arxiv.org/abs/2004.12832). (token 级 MaxSim 晚交互与可预计算文档表示)

*   **[Malkov & Yashunin, 2018]** [*Efficient and Robust Approximate Nearest Neighbor Search Using Hierarchical Navigable Small World Graphs*](https://arxiv.org/abs/1603.09320). (HNSW 图索引与近似最近邻的延迟—召回折中)

*   **[Liu et al., 2023]** [*Lost in the Middle: How Language Models Use Long Contexts*](https://arxiv.org/abs/2307.03172). (多文档问答与键值检索中的证据位置效应；结论受模型和任务协议约束)

*   **[Model Context Protocol, 2025-11-25]** [*Specification*](https://modelcontextprotocol.io/specification/2025-11-25). (模型应用与工具、资源、提示模板等上下文集成；`latest` 核验于 2026-07-12)

*   **[A2A Project, 2026]** [*A2A Protocol Specification v1.0.0*](https://a2a-protocol.org/latest/specification/). (2026 年首个 stable 规范：Agent 间消息、任务与 artifact 协调；Google 发起，现为 Linux Foundation 项目；核验于 2026-07-12)
*   **[OpenAI, 核验于 2026-07-12]** [*OpenAI Agents SDK Documentation*](https://openai.github.io/openai-agents-python/). (Agent 运行时、工具、handoff、guardrails 与 tracing 的实现文档)
*   **[LangChain, 核验于 2026-07-12]** [*LangGraph Overview*](https://docs.langchain.com/oss/python/langgraph/overview). (持久化执行、状态图与多 Agent 编排的实现文档)
*   **[Google, 核验于 2026-07-12]** [*Agent Development Kit Documentation*](https://google.github.io/adk-docs/). (Agent 开发、会话、记忆、工具与上下文管理的实现文档)

## 7. 推荐书籍 (Recommended Books)

*   **Ian Goodfellow, Yoshua Bengio, Aaron Courville**. *Deep Learning*. MIT Press, 2016. (深度学习教材)
*   **Sutton & Barto**. *Reinforcement Learning: An Introduction*. MIT Press, 2018. (强化学习入门)
*   **Daniel Kahneman**. *Thinking, Fast and Slow*. (对 System 1 / System 2 表述的普及；双过程理论来自更广泛研究传统)