Feature Attribution papers, 2023–2026

Key findings

• Feature Attribution went from 118 papers in 2023 to 219 in 2026, roughly 1.9x in three years.
• NeurIPS publishes the most Feature Attribution work: 193 papers from 2023 to 2026, 28% of the four-venue total.
• The fastest-growing overlap is with Theory · learning theory: papers tagged with both rose from 1 in 2023 to 14 in 2026.
• Among the 5 topics under Interpretability, it ranks #1 by 2026 paper count; the fastest grower in the group, Mechanistic Interpretability, is up 8.1x versus 1.9x here.

Representative papers

Spread across years and venues so every edition shows up

• NeurIPS 2026ADAPT: Hybrid Prompt Optimization for LLM Feature Visualization
• ICML 2026Activation with Intrinsic-Extrinsic Consensus
• ICLR 2026AnyUp: Universal Feature Upsampling
• CVPR 2026A Training-Free Style-Personalization via SVD-Based Feature Decomposition
• NeurIPS 2025Alternating Gradient Flows: A Theory of Feature Learning in Two-layer Neural Networks
• ICML 2025BCE vs. CE in Deep Feature Learning
• ICLR 2025A Differentiable Rank-Based Objective for Better Feature Learning
• CVPR 2025A Simple Data Augmentation for Feature Distribution Skewed Federated Learning
• NeurIPS 2024$\texttt{dattri}$: A Library for Efficient Data Attribution
• ICML 2024Asymptotics of feature learning in two-layer networks after one gradient-step
• ICLR 2024An Intuitive Multi-Frequency Feature Representation for SO(3)-Equivariant Networks
• CVPR 2024DYSON: Dynamic Feature Space Self-Organization for Online Task-Free Class Incremental Learning
• NeurIPS 2023$\mathcal{M}^4$: A Unified XAI Benchmark for Faithfulness Evaluation of Feature Attribution Methods across Metrics, Modalities and Models
• ICML 2023Decoding Layer Saliency in Language Transformers
• ICLR 2023Can We Faithfully Represent Absence States to Compute Shapley Values on a DNN?
• CVPR 2023A Practical Upper Bound for the Worst-Case Attribution Deviations
• NeurIPS 2026Adaptive Feature Propagation for Attribute-Missing Graph Clustering
• ICML 2026Aggregate Models, Not Explanations: Improving Feature Importance Estimation
• ICLR 2026Attribution-Guided Decoding
• CVPR 2026FBTA Enabling Single-GPU End-to-End Gigapixel WSI Classification with Feature Bridging

All Feature Attribution papers

175 papers, page 1 of 4

1. ICML 2026$\texttt{ShaplEIG}$: Bayesian Experimental Design for Shapley Value Estimation
2. ICML 2026A unified theory of feature learning in RNNs and DNNs
3. ICML 2026Activation with Intrinsic-Extrinsic Consensus
4. ICML 2026Aggregate Models, Not Explanations: Improving Feature Importance Estimation
5. ICML 2026Alignment-Sensitive Minimax Rates for Spectral Algorithms with Learned Kernels
6. ICML 2026Barriers to Counterfactual Credit Attribution for Autoregressive Models
7. ICML 2026BASIL: Scalable Bayesian Semi-supervised Clustering with Feature Selection and Adaptive Constraint Weighting
8. ICML 2026Beyond Trajectory-Level Attribution: Graph-Based Credit Assignment for Agentic Reinforcement Learning
9. ICML 2026C$^{2}$R: Cross-sample Consistency Regularization Mitigates Feature Splitting and Absorption in Sparse Autoencoders
10. ICML 2026Causal Feature Learning via Generalized Rayleigh Quotients
11. ICML 2026ConEx: Human-Interpretable Saliency Maps via Concept-Aware Attribution
12. ICML 2026DAVE: Distribution-Aware Attribution via ViT Gradient Decomposition
13. ICML 2026Decomposing Query-Key Feature Interactions Using Contrastive Covariances
14. ICML 2026Detecting and Filtering Unsafe Training Data via Data Attribution with Denoised Representation
15. ICML 2026Dichotomy of Feature Learning and Unlearning: Fast-Slow Analysis on Neural Networks with Stochastic Gradient Descent
16. ICML 2026Diversity-Aware Recursive Feature Multiple Kernel Learning
17. ICML 2026Expand Neurons, Not Parameters
18. ICML 2026ExPLAIND: Unifying Model, Data, and Training Attribution to Study Model Behavior
19. ICML 2026Explaining Concept Shift with Interpretable Feature Attribution
20. ICML 2026Feature Bagging Provides Stability
21. ICML 2026Feature Collapse Under Corruption: An Entropy Perspective on Robust Neural Networks
22. ICML 2026Feature Resemblance: Towards a Theoretical Understanding of Analogical Reasoning in Transformers
23. ICML 2026From Parameters to Feature Space: Task Arithmetic for Backdoor Mitigation in Model Merging
24. ICML 2026GFD-EMVC: Evolutionary Multi-View Classification with Label Noise via Gradient and Feature Dual-Perception
25. ICML 2026GOTabPFN: From Feature Ordering to Compact Tokenization for Tabular Foundation Models on High-Dimensional Data
26. ICML 2026Gradient Flow Through Diagram Expansions: Learning Regimes and Explicit Solutions
27. ICML 2026GUDA: Counterfactual Group-wise Training Data Attribution for Diffusion Models via Unlearning
28. ICML 2026Hidden in Plain Sight -- Class Competition Focuses Attribution Maps
29. ICML 2026Ideal Attribution and Faithful Watermarks for Language Models
30. ICML 2026Improving Adversarial Robustness of Attribution via Implicit Regularization
31. ICML 2026Learning-To-Measure: In-Context Active Feature Acquisition
32. ICML 2026Less is Enough: Synthesizing Diverse Data in LLM Feature Space with Sparse Autoencoders
33. ICML 2026Manifold-Aligned Guided Integrated Gradients for Reliable Feature Attribution
34. ICML 2026Mechanistic Data Attribution: Tracing the Training Origins of Interpretable LLM Units
35. ICML 2026Neural Feature Geometry Evolves as Discrete Ricci Flow
36. ICML 2026Nonparametric Data Attribution for Diffusion Models
37. ICML 2026On the Accuracy of Newton Step and Influence Function Data Attributions
38. ICML 2026On the Power of Source Screening for Learning Shared Feature Extractors
39. ICML 2026On the Relationship Between Activation Outliers and Feature Death in Sparse Autoencoders
40. ICML 2026Over-Alignment vs Over-Fitting: The Role of Feature Learning Strength in Generalization
41. ICML 2026PACEAttention: Principled and Adaptive Feature Compression-Expansion Grounded in the Geometry of $\text{MCR}^2$
42. ICML 2026PLASH: Provably Linear-Time Attention with Selective Higher-Order Feature Sketching
43. ICML 2026Priority-Aware Shapley Value
44. ICML 2026QPoint: End-to-End Lightweight Point Cloud Processing via Robust Quaternion Feature Learning
45. ICML 2026Resolution as a Direction: Vector-Panning Feature Alignment for Cross-Resolution Re-Identification
46. ICML 2026Robustness of Mixtures of Experts to Feature Noise
47. ICML 2026SafeSeek: Universal Attribution of Safety Circuits in Language Models
48. ICML 2026Shapley Regularized Neural Granger Causality
49. ICML 2026Shared Semantics, Divergent Mechanisms: Unsupervised Feature Discovery by Aligning Semantics and Mechanisms
50. ICML 2026SSA: Sparse Sparse Attention by Aligning Full and Sparse Attention Outputs in Feature Space

FAQ

How many Feature Attribution papers were published at NeurIPS, ICLR, ICML and CVPR from 2023 to 2026?

695 in total: 118 in 2023, 168 in 2024, 190 in 2025, 219 in 2026.

How many Feature Attribution papers are at NeurIPS 2026?

NeurIPS 2026 has 61 Feature Attribution papers, versus 57 the year before.

How many Feature Attribution papers are at ICLR 2026?

ICLR 2026 has 53 Feature Attribution papers, versus 42 the year before.

How many Feature Attribution papers are at ICML 2026?

ICML 2026 has 56 Feature Attribution papers, versus 51 the year before.

How many Feature Attribution papers are at CVPR 2026?

CVPR 2026 has 49 Feature Attribution papers, versus 40 the year before.

Which conference publishes the most Feature Attribution papers?

NeurIPS, with 193 papers from 2023 to 2026.