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 2 / 2 页

51. NeurIPS 2024Algorithmic Capabilities of Random Transformers
52. NeurIPS 2024An Analysis of Tokenization: Transformers under Markov Data
53. NeurIPS 2024B-cosification: Transforming Deep Neural Networks to be Inherently Interpretable
54. NeurIPS 2024Compact Proofs of Model Performance via Mechanistic Interpretability
55. NeurIPS 2024Confidence Regulation Neurons in Language Models
56. NeurIPS 2024CoSy: Evaluating Textual Explanations of Neurons
57. NeurIPS 2024Emergence of Hidden Capabilities: Exploring Learning Dynamics in Concept Space
58. NeurIPS 2024Evidence of Learned Look-Ahead in a Chess-Playing Neural Network
59. NeurIPS 2024Grokking of Implicit Reasoning in Transformers: A Mechanistic Journey to the Edge of Generalization
60. NeurIPS 2024Hypothesis Testing the Circuit Hypothesis in LLMs
61. NeurIPS 2024Improving Sparse Decomposition of Language Model Activations with Gated Sparse Autoencoders
62. NeurIPS 2024InterpBench: Semi-Synthetic Transformers for Evaluating Mechanistic Interpretability Techniques
63. NeurIPS 2024InversionView: A General-Purpose Method for Reading Information from Neural Activations
64. NeurIPS 2024Iteration Head: A Mechanistic Study of Chain-of-Thought
65. NeurIPS 2024Linking In-context Learning in Transformers to Human Episodic Memory
66. NeurIPS 2024LLM Circuit Analyses Are Consistent Across Training and Scale
67. NeurIPS 2024Measuring Per-Unit Interpretability at Scale Without Humans
68. NeurIPS 2024Measuring Progress in Dictionary Learning for Language Model Interpretability with Board Game Models
69. NeurIPS 2024Optimal ablation for interpretability
70. NeurIPS 2024Partial observation can induce mechanistic mismatches in data-constrained models of neural dynamics
71. NeurIPS 2024Refusal in Language Models Is Mediated by a Single Direction
72. NeurIPS 2024Talking Heads: Understanding Inter-Layer Communication in Transformer Language Models
73. NeurIPS 2024Transcoders find interpretable LLM feature circuits
74. NeurIPS 2024Truth is Universal: Robust Detection of Lies in LLMs
75. NeurIPS 2024What Makes and Breaks Safety Fine-tuning? A Mechanistic Study
76. NeurIPS 2023A Logic for Expressing Log-Precision Transformers
77. NeurIPS 2023Analyzing Vision Transformers for Image Classification in Class Embedding Space
78. NeurIPS 2023Concentration analysis of multivariate elliptic diffusions
79. NeurIPS 2023How does GPT-2 compute greater-than?: Interpreting mathematical abilities in a pre-trained language model
80. NeurIPS 2023Interpretability at Scale: Identifying Causal Mechanisms in Alpaca
81. NeurIPS 2023Learning Transformer Programs
82. NeurIPS 2023SmoothHess: ReLU Network Feature Interactions via Stein's Lemma
83. NeurIPS 2023Towards Automated Circuit Discovery for Mechanistic Interpretability
84. NeurIPS 2023Tracr: Compiled Transformers as a Laboratory for Interpretability
85. NeurIPS 2023YouTubePD: A Multimodal Benchmark for Parkinson’s Disease Analysis

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。