概念解释论文趋势 2023–2026

主要结论

• 概念解释论文从 2023 年的 91 篇增长到 2026 年的 117 篇，三年间扩大到约 1.3 倍。
• NeurIPS 是发概念解释论文最多的会议，2023–2026 年共 134 篇，占四个会议总量的 39%。
• 交叉方向里涨得最快的是多模态大模型：同时属于两个主题的论文从 2023 年的 4 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 3；同组增长最快的机制解释是 8.1 倍，概念解释是 1.3 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 2026CENDRe: Concept Extraction with Natural Domain Representations
• ICML 2026Evidential Copula Concept Embedding Models
• ICLR 2026Decoupling the Class Label and the Target Concept in Machine Unlearning
• CVPR 2026Agile Deliberation Concept Deliberation for Subjective Visual Classification
• NeurIPS 2025An Analysis of Concept Bottleneck Models: Measuring, Understanding, and Mitigating the Impact of Noisy Annotations
• ICML 2025Concept-Centric Token Interpretation for Vector-Quantized Generative Models
• ICLR 2025Causal Concept Graph Models: Beyond Causal Opacity in Deep Learning
• CVPR 2025Discovering Fine-Grained Visual-Concept Relations by Disentangled Optimal Transport Concept Bottleneck Models
• NeurIPS 2024[Re] On the Reproducibility of Post-Hoc Concept Bottleneck Models
• ICML 2024ContPhy: Continuum Physical Concept Learning and Reasoning from Videos
• ICLR 2024AttEXplore: Attribution for Explanation with model parameters eXploration
• CVPR 2024Fine-Grained Bipartite Concept Factorization for Clustering
• NeurIPS 2023[Re] Hierarchical Shrinkage: Improving the Accuracy and Interpretability of Tree-Based Methods
• ICML 2023A Closer Look at the Intervention Procedure of Concept Bottleneck Models
• ICLR 2023Concept Gradient: Concept-based Interpretation Without Linear Assumption
• CVPR 2023CRAFT: Concept Recursive Activation FacTorization for Explainability
• NeurIPS 2026Concept frustration: Aligning human concepts and machine representations
• ICML 2026On the Variability of Concept Activation Vectors
• ICLR 2026Hierarchical Concept-based Interpretable Models
• CVPR 2026DCoAR Deep Concept Injection into Unified Autoregressive Models for Personalized

全部概念解释论文

共 342 篇，第 1 / 7 页

1. CVPR 2026Agile Deliberation Concept Deliberation for Subjective Visual Classification
2. CVPR 2026AXG-Reasoner Error Detection and Explanation in Long Task Videos with
3. CVPR 2026CLIP-like Model as a Foundational Density Ratio Estimator
4. CVPR 2026Closed-Form Concept Erasure via Double Projections
5. CVPR 2026Composing Concepts from Images and Videos via Concept-prompt Binding
6. CVPR 2026Concept Regions Matter Benchmarking CLIP with a New Cluster-Importance Approach
7. CVPR 2026Concept-Guided Fine-Tuning Steering ViTs away from Spurious Correlations to Improve
8. CVPR 2026ConceptPrism Concept Disentanglement in Personalized Diffusion Models via Residual Token
9. CVPR 2026CUE Concept-Aware Multi-Label Expansion to Mitigate Concept Confusion in Long-Tailed
10. CVPR 2026DCoAR Deep Concept Injection into Unified Autoregressive Models for Personalized
11. CVPR 2026Dictionary-Aligned Concept Control for Safeguarding Multimodal LLMs
12. CVPR 2026DSCA Dynamic Subspace Concept Alignment for Lifelong VLM Editing
13. CVPR 2026EMMA Concept Erasure Benchmark with Comprehensive Semantic Metrics and Diverse
14. CVPR 2026Erasing Thousands of Concepts Towards Scalable and Practical Concept Erasure
15. CVPR 2026GenErase Generalizable and Semantically-Aware Concept Erasure in Diffusion Models
16. CVPR 2026Hierarchical Concept Embedding Pursuit for Interpretable Image Classification
17. CVPR 2026Interpretable and Steerable Concept Bottleneck Sparse Autoencoders
18. CVPR 2026Intrinsic Concept Extraction Based on Compositional Interpretability
19. CVPR 2026Learning Like Humans Analogical Concept Learning for Generalized Category Discovery
20. CVPR 2026Learning What Matters Prioritized Concept Learning via Relative Error-driven Sample
21. CVPR 2026Making the Classification Explanation Faithful to the Confidence Score
22. CVPR 2026MapRoutePrecise-Concept Erasing Mappers via Semantic Routing
23. CVPR 2026Measuring the UnFaithfulness of Concept-Based Explanations
24. CVPR 2026No Hard Negatives Required Concept Centric Learning Leads to Compositionality
25. CVPR 2026Omni-Attribute Open-vocabulary Attribute Encoder for Visual Concept Personalization
26. CVPR 2026OrthoFuse Training-free Riemannian Fusion of Orthogonal Style-Concept Adapters for Diffusion
27. CVPR 2026Prototype-Guided Concept Erasure in Diffusion Models
28. CVPR 2026Rethinking Concept Bottleneck Models From Pitfalls to Solutions
29. CVPR 2026Roots Beneath the Cut Uncovering the Risk of Concept Revival
30. CVPR 2026Rounded or Streamlined Head Bridging Concept Bottleneck Models and Attribute-Described
31. CVPR 2026Select Hypothesize and Verify Towards Verified Neuron Concept Interpretation
32. CVPR 2026TokenTrace Multi-Concept Attribution through Watermarked Token Recovery
33. CVPR 2026UniVerse A Unified Modulation Framework for Segmentation-Free Disentangled Multi-Concept Personalization
34. ICLR 2026A Probabilistic Hard Concept Bottleneck for Steerable Generative Models
35. ICLR 2026CASteer: Cross-Attention Steering for Controllable Concept Erasure
36. ICLR 2026Causality ≠ Invariance: Function and Concept Vectors in LLMs
37. ICLR 2026Closing the Safety Gap: Surgical Concept Erasure in Visual Autoregressive Models
38. ICLR 2026Co-occurring Associated REtained concepts in Diffusion Unlearning
39. ICLR 2026Concept-based Adversarial Attack: a Probabilistic Perspective
40. ICLR 2026Debugging Concept Bottleneck Models through Removal and Retraining
41. ICLR 2026Decomposition of Concept-Level Rules in Visual Scenes
42. ICLR 2026Decoupling the Class Label and the Target Concept in Machine Unlearning
43. ICLR 2026Evaluating SAE interpretability without generating explanations
44. ICLR 2026Forget Many, Forget Right: Scalable and Precise Concept Unlearning in Diffusion Models
45. ICLR 2026Hierarchical Concept-based Interpretable Models
46. ICLR 2026LAMDA: A Longitudinal Android Malware Benchmark for Concept Drift Analysis
47. ICLR 2026Latent Concept Disentanglement in Transformer-based Language Models
48. ICLR 2026Leveraging Explanation to Improve Generalization of Meta Reinforcement Learning
49. ICLR 2026RepIt: Steering Language Models with Concept-Specific Refusal Vectors
50. ICLR 2026Sparling: End-to-End Spatial Concept Learning via Extremely Sparse Activations

常见问题

2023–2026 年四大会议一共有多少篇概念解释论文？

共 342 篇，按年份分别是 2023 年 91 篇、2024 年 63 篇、2025 年 71 篇、2026 年 117 篇。

NeurIPS 2026 有多少篇概念解释论文？

NeurIPS 2026 有 40 篇概念解释论文，上一届是 25 篇。

ICLR 2026 有多少篇概念解释论文？

ICLR 2026 有 24 篇概念解释论文，上一届是 19 篇。

ICML 2026 有多少篇概念解释论文？

ICML 2026 有 20 篇概念解释论文，上一届是 11 篇。

CVPR 2026 有多少篇概念解释论文？

CVPR 2026 有 33 篇概念解释论文，上一届是 16 篇。

哪个会议的概念解释论文最多？

NeurIPS，2023–2026 年共 134 篇。