机制解释论文趋势 2023–2026

电路分析

统计范围：NeurIPS、ICLR、ICML、CVPR 四个会议，2023–2026 年，共 311 篇被归入该主题的论文。

主要结论

每年论文数

各会议论文数

按年份和会议统计

机制解释论文数，按年份和会议
年份NeurIPSICLRICMLCVPR合计
20231043017
202425149149
20253438332107
20261646742138

代表论文

按年份和会议均匀挑选，方便快速看到每一届的样貌

全部机制解释论文

年份全部2023202420252026
会议全部NeurIPSICLRICMLCVPR

共 311 篇，第 5 / 7 页

  1. ICML 2025Scaling Sparse Feature Circuits For Studying In-Context Learning
  2. ICML 2025Strategy Coopetition Explains the Emergence and Transience of In-Context Learning
  3. ICML 2025Taming Knowledge Conflicts in Language Models
  4. ICML 2025The Logical Implication Steering Method for Conditional Interventions on Transformer Generation
  5. ICML 2025The underlying structures of self-attention: symmetry, directionality, and emergent dynamics in Transformer training
  6. ICML 2025To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models
  7. ICML 2025Towards Global-level Mechanistic Interpretability: A Perspective of Modular Circuits of Large Language Models
  8. ICML 2025Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models
  9. ICML 2025Towards Understanding Fine-Tuning Mechanisms of LLMs via Circuit Analysis
  10. ICML 2025Validating Mechanistic Interpretations: An Axiomatic Approach
  11. ICML 2025Which Attention Heads Matter for In-Context Learning?
  12. NeurIPS 2025A Implies B: Circuit Analysis in LLMs for Propositional Logical Reasoning
  13. NeurIPS 2025Angular Steering: Behavior Control via Rotation in Activation Space
  14. NeurIPS 2025Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings
  15. NeurIPS 2025Better Training Data Attribution via Better Inverse Hessian-Vector Products
  16. NeurIPS 2025Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits
  17. NeurIPS 2025Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models
  18. NeurIPS 2025Capturing Polysemanticity with PRISM: A Multi-Concept Feature Description Framework
  19. NeurIPS 2025Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers
  20. NeurIPS 2025Dense SAE Latents Are Features, Not Bugs
  21. NeurIPS 2025DePass: Unified Feature Attributing by Simple Decomposed Forward Pass
  22. NeurIPS 2025Detecting High-Stakes Interactions with Activation Probes
  23. NeurIPS 2025Do different prompting methods yield a common task representation in language models?
  24. NeurIPS 2025EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification
  25. NeurIPS 2025FaCT: Faithful Concept Traces for Explaining Neural Network Decisions
  26. NeurIPS 2025Failure by Interference: Language Models Make Balanced Parentheses Errors When Faulty Mechanisms Overshadow Sound Ones
  27. NeurIPS 2025Improved Representation Steering for Language Models
  28. NeurIPS 2025Interpreting Emergent Features in Deep Learning-based Side-channel Analysis
  29. NeurIPS 2025Interpreting vision transformers via residual replacement model
  30. NeurIPS 2025LLM Interpretability with Identifiable Temporal-Instantaneous Representation
  31. NeurIPS 2025LLM Layers Immediately Correct Each Other
  32. NeurIPS 2025Manipulating Feature Visualizations with Gradient Slingshots
  33. NeurIPS 2025Mechanistic Interpretability of RNNs emulating Hidden Markov Models
  34. NeurIPS 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steering
  35. NeurIPS 2025Pinpointing Attention-Causal Communication in Language Models
  36. NeurIPS 2025Prompting as Scientific Inquiry
  37. NeurIPS 2025Remarkable Robustness of LLMs: Stages of Inference?
  38. NeurIPS 2025Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates
  39. NeurIPS 2025Revising and Falsifying Sparse Autoencoder Feature Explanations
  40. NeurIPS 2025Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs
  41. NeurIPS 2025The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?
  42. NeurIPS 2025The Structure of Relation Decoding Linear Operators in Large Language Models
  43. NeurIPS 2025Uncovering a Universal Abstract Algorithm for Modular Addition in Neural Networks
  44. NeurIPS 2025Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning
  45. NeurIPS 2025What One Cannot, Two Can: Two-Layer Transformers Provably Represent Induction Heads on Any-Order Markov Chains
  46. CVPR 2024SI-MIL: Taming Deep MIL for Self-Interpretability in Gigapixel Histopathology
  47. ICLR 2024Attention Satisfies: A Constraint-Satisfaction Lens on Factual Errors of Language Models
  48. ICLR 2024Circuit Component Reuse Across Tasks in Transformer Language Models
  49. ICLR 2024Feature emergence via margin maximization: case studies in algebraic tasks
  50. ICLR 2024Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking

在 Claude、Cursor 里直接问这个趋势

接上 Paperlist MCP 后，写综述、挑方向或准备投稿时，不用离开对话框就能拿到这页的数字：每年各会议的论文数、增长倍数、涨得最快的子方向和代表论文，AI 会按你的问题整理。

可以直接复制这句话问 AI

用 paperlist 查机制解释在 ICML 2023–2026 的论文趋势，哪一年涨得最快？

AI 会调用的工具：get_topic_stats {"topic": "mechanistic-interpretability"}

两分钟接入 Paperlist MCP（Claude Code、Cursor、Codex、ChatGPT）

常见问题

2023–2026 年四大会议一共有多少篇机制解释论文？
共 311 篇，按年份分别是 2023 年 17 篇、2024 年 49 篇、2025 年 107 篇、2026 年 138 篇。
NeurIPS 2026 有多少篇机制解释论文？
NeurIPS 2026 有 16 篇机制解释论文，上一届是 34 篇。
ICLR 2026 有多少篇机制解释论文？
ICLR 2026 有 46 篇机制解释论文，上一届是 38 篇。
ICML 2026 有多少篇机制解释论文？
ICML 2026 有 74 篇机制解释论文，上一届是 33 篇。
CVPR 2026 有多少篇机制解释论文？
CVPR 2026 有 2 篇机制解释论文，上一届是 2 篇。
哪个会议的机制解释论文最多？
ICML，2023–2026 年共 119 篇。

「可解释性」下的其他主题