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 107 篇，第 2 / 3 页

51. ICML 2025In-Context Linear Regression Demystified: Training Dynamics and Mechanistic Interpretability of Multi-Head Softmax Attention
52. ICML 2025Inference-Time Decomposition of Activations (ITDA): A Scalable Approach to Interpreting Large Language Models
53. ICML 2025Learning Multi-Level Features with Matryoshka Sparse Autoencoders
54. ICML 2025Locate-then-edit for Multi-hop Factual Recall under Knowledge Editing
55. ICML 2025Looking Beyond the Top-1: Transformers Determine Top Tokens in Order
56. ICML 2025Mechanistic Unlearning: Robust Knowledge Unlearning and Editing via Mechanistic Localization
57. ICML 2025MIB: A Mechanistic Interpretability Benchmark
58. ICML 2025Models of Heavy-Tailed Mechanistic Universality
59. ICML 2025Position: Explainable AI Cannot Advance Without Better User Studies
60. ICML 2025Position: Rethinking Explainable Machine Learning as Applied Statistics
61. ICML 2025Representation Shattering in Transformers: A Synthetic Study with Knowledge Editing
62. ICML 2025SAEBench: A Comprehensive Benchmark for Sparse Autoencoders in Language Model Interpretability
63. ICML 2025Scaling Sparse Feature Circuits For Studying In-Context Learning
64. ICML 2025Strategy Coopetition Explains the Emergence and Transience of In-Context Learning
65. ICML 2025Taming Knowledge Conflicts in Language Models
66. ICML 2025The Logical Implication Steering Method for Conditional Interventions on Transformer Generation
67. ICML 2025The underlying structures of self-attention: symmetry, directionality, and emergent dynamics in Transformer training
68. ICML 2025To Steer or Not to Steer? Mechanistic Error Reduction with Abstention for Language Models
69. ICML 2025Towards Global-level Mechanistic Interpretability: A Perspective of Modular Circuits of Large Language Models
70. ICML 2025Towards scientific discovery with dictionary learning: Extracting biological concepts from microscopy foundation models
71. ICML 2025Towards Understanding Fine-Tuning Mechanisms of LLMs via Circuit Analysis
72. ICML 2025Validating Mechanistic Interpretations: An Axiomatic Approach
73. ICML 2025Which Attention Heads Matter for In-Context Learning?
74. NeurIPS 2025A Implies B: Circuit Analysis in LLMs for Propositional Logical Reasoning
75. NeurIPS 2025Angular Steering: Behavior Control via Rotation in Activation Space
76. NeurIPS 2025Attention Sinks: A 'Catch, Tag, Release' Mechanism for Embeddings
77. NeurIPS 2025Better Training Data Attribution via Better Inverse Hessian-Vector Products
78. NeurIPS 2025Beyond Components: Singular Vector-Based Interpretability of Transformer Circuits
79. NeurIPS 2025Bigram Subnetworks: Mapping to Next Tokens in Transformer Language Models
80. NeurIPS 2025Capturing Polysemanticity with PRISM: A Multi-Concept Feature Description Framework
81. NeurIPS 2025Causal Head Gating: A Framework for Interpreting Roles of Attention Heads in Transformers
82. NeurIPS 2025Dense SAE Latents Are Features, Not Bugs
83. NeurIPS 2025DePass: Unified Feature Attributing by Simple Decomposed Forward Pass
84. NeurIPS 2025Detecting High-Stakes Interactions with Activation Probes
85. NeurIPS 2025Do different prompting methods yield a common task representation in language models?
86. NeurIPS 2025EAP-GP: Mitigating Saturation Effect in Gradient-based Automated Circuit Identification
87. NeurIPS 2025FaCT: Faithful Concept Traces for Explaining Neural Network Decisions
88. NeurIPS 2025Failure by Interference: Language Models Make Balanced Parentheses Errors When Faulty Mechanisms Overshadow Sound Ones
89. NeurIPS 2025Improved Representation Steering for Language Models
90. NeurIPS 2025Interpreting Emergent Features in Deep Learning-based Side-channel Analysis
91. NeurIPS 2025Interpreting vision transformers via residual replacement model
92. NeurIPS 2025LLM Interpretability with Identifiable Temporal-Instantaneous Representation
93. NeurIPS 2025LLM Layers Immediately Correct Each Other
94. NeurIPS 2025Manipulating Feature Visualizations with Gradient Slingshots
95. NeurIPS 2025Mechanistic Interpretability of RNNs emulating Hidden Markov Models
96. NeurIPS 2025Mitigating Overthinking in Large Reasoning Models via Manifold Steering
97. NeurIPS 2025Pinpointing Attention-Causal Communication in Language Models
98. NeurIPS 2025Prompting as Scientific Inquiry
99. NeurIPS 2025Remarkable Robustness of LLMs: Stages of Inference?
100. NeurIPS 2025Rethinking Circuit Completeness in Language Models: AND, OR, and ADDER Gates

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。