机制解释论文趋势 2023–2026

主要结论

• 机制解释论文从 2023 年的 17 篇增长到 2026 年的 138 篇，三年间扩大到约 8.1 倍。
• ICML 是发机制解释论文最多的会议，2023–2026 年共 119 篇，占四个会议总量的 38%。
• 交叉方向里涨得最快的是深度学习 · large language models：同时属于两个主题的论文从 2023 年的 1 篇增至 2026 年的 23 篇。
• 在「可解释性」的 5 个主题里，它的 2026 年论文数排第 2；同组增长最快的特征归因是 1.9 倍，机制解释是 8.1 倍。

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

• NeurIPS 2026A Mechanistic Interpretability Study of an Astronomical Foundation Model
• ICML 2026Contribution Weights: A Geometrical Analysis of Self-Attention Transformers
• ICLR 2026Circuit Insights: Towards Interpretability Beyond Activations
• CVPR 2026CIGMA Causal Information-Gain Mechanistic Attribution of Attention Heads in Vision
• NeurIPS 2025Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits
• ICML 2025Algorithm Development in Neural Networks: Insights from the Streaming Parity Task
• ICLR 2025Arithmetic Without Algorithms: Language Models Solve Math with a Bag of Heuristics
• CVPR 2025Dissecting and Mitigating Diffusion Bias via Mechanistic Interpretability
• NeurIPS 2024InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques
• ICML 2024Mechanistic Neural Networks for Scientific Machine Learning
• ICLR 2024Fine-Tuning Enhances Existing Mechanisms: A Case Study on Entity Tracking
• CVPR 2024SI-MIL: Taming Deep MIL for Self-Interpretability in Gigapixel Histopathology
• NeurIPS 2023A Logic for Expressing Log-Precision Transformers
• ICML 2023Mechanistic Mode Connectivity
• ICLR 2023Interpretability in the Wild: a Circuit for Indirect Object Identification in GPT-2 Small
• NeurIPS 2026Discovering Mechanistic Models of Neural Activity: System Identification in an in Silico Zebrafish
• ICML 2026CorrSteer: Generation-Time LLM Steering via Correlated Sparse Autoencoder Features
• ICLR 2026CLUE: Conflict-guided Localization for LLM Unlearning Framework
• CVPR 2026From Weights to Concepts Data-Free Interpretability of CLIP via Singular
• NeurIPS 2025Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models

全部机制解释论文

共 85 篇，第 1 / 2 页

1. NeurIPS 2026A Mechanistic Interpretability Study of an Astronomical Foundation Model
2. NeurIPS 2026A Unifying Perspective on Language Model Interpretability
3. NeurIPS 2026Assign and Add: A Mechanistic Study of Compositional Arithmetic
4. NeurIPS 2026Circuit-Level Knowledge Distillation for Large Language Models
5. NeurIPS 2026Discovering Mechanistic Models of Neural Activity: System Identification in an in Silico Zebrafish
6. NeurIPS 2026Geometry-Adaptive Explainer for Faithful Dictionary-Based Interpretability under Distribution Shift
7. NeurIPS 2026How Do Language Models Understand Tables? A Mechanistic Analysis of Cell Location
8. NeurIPS 2026How do Small Transformer Models Learn Hard Math Tasks?
9. NeurIPS 2026Instruction Anchor: Dissecting the Mechanistic Dynamics of Modality Arbitration
10. NeurIPS 2026OceanCBM: A Concept Bottleneck Model for Mechanistic Interpretability in Ocean Forecasting
11. NeurIPS 2026On the Recall Scaling Laws in Mamba: A Theoretical and Mechanistic Study via Hashing
12. NeurIPS 2026Patching Up Circuit Learning on Images
13. NeurIPS 2026Predictive Concept Decoders: Training Scalable End-to-End Interpretability Assistants
14. NeurIPS 2026Structural Support Certificates for Mechanistic Hypothesis Selection
15. NeurIPS 2026The parameters in weight-sparse transformers are interpretable
16. NeurIPS 2026The Unembedding Bottleneck: A Mechanistic Analysis of Single Digit Counting in LLMs
17. NeurIPS 2025A Implies B: Circuit Analysis in LLMs for Propositional Logical Reasoning
18. NeurIPS 2025Angular Steering: Behavior Control via Rotation in Activation Space
19. NeurIPS 2025Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings
20. NeurIPS 2025Better Training Data Attribution via Better Inverse Hessian-Vector Products
21. NeurIPS 2025Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits
22. NeurIPS 2025Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models
23. NeurIPS 2025Capturing Polysemanticity with PRISM: A Multi-Concept Feature Description Framework
24. NeurIPS 2025Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers
25. NeurIPS 2025Dense SAE Latents Are Features, Not Bugs
26. NeurIPS 2025DePass: Unified Feature Attributing by Simple Decomposed Forward Pass
27. NeurIPS 2025Detecting High-Stakes Interactions with Activation Probes
28. NeurIPS 2025Do different prompting methods yield a common task representation in language models?
29. NeurIPS 2025EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification
30. NeurIPS 2025FaCT: Faithful Concept Traces for Explaining Neural Network Decisions
31. NeurIPS 2025Failure by Interference: Language Models Make Balanced Parentheses Errors When Faulty Mechanisms Overshadow Sound Ones
32. NeurIPS 2025Improved Representation Steering for Language Models
33. NeurIPS 2025Interpreting Emergent Features in Deep Learning-based Side-channel Analysis
34. NeurIPS 2025Interpreting vision transformers via residual replacement model
35. NeurIPS 2025LLM Interpretability with Identifiable Temporal-Instantaneous Representation
36. NeurIPS 2025LLM Layers Immediately Correct Each Other
37. NeurIPS 2025Manipulating Feature Visualizations with Gradient Slingshots
38. NeurIPS 2025Mechanistic Interpretability of RNNs emulating Hidden Markov Models
39. NeurIPS 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steering
40. NeurIPS 2025Pinpointing Attention-Causal Communication in Language Models
41. NeurIPS 2025Prompting as Scientific Inquiry
42. NeurIPS 2025Remarkable Robustness of LLMs: Stages of Inference?
43. NeurIPS 2025Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates
44. NeurIPS 2025Revising and Falsifying Sparse Autoencoder Feature Explanations
45. NeurIPS 2025Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs
46. NeurIPS 2025The Non-Linear Representation Dilemma: Is Causal Abstraction Enough for Mechanistic Interpretability?
47. NeurIPS 2025The Structure of Relation Decoding Linear Operators in Large Language Models
48. NeurIPS 2025Uncovering a Universal Abstract Algorithm for Modular Addition in Neural Networks
49. NeurIPS 2025Unifying Attention Heads and Task Vectors via Hidden State Geometry in In-Context Learning
50. NeurIPS 2025What One Cannot, Two Can: Two-Layer Transformers Provably Represent Induction Heads on Any-Order Markov Chains

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。