Research map / Label-efficient and robust learning / Recognition and long-tail
Understanding Learned Models by Identifying Important Features at the Right Resolution
Kyubin Lee, Akshay Sood, Mark Craven
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Increase the synthetic test-set size from m=32 to m=16384 | Proposed feature and interaction testing method · Synthetic data | feature power: 0.722 -> 0.975 | Improved |
| Increase the synthetic test-set size from m=32 to m=16384 | Proposed feature and interaction testing method · Synthetic data | interaction power: 0.132 -> 0.949 | Improved |
| Increase model-output noise from σ=0.00 to σ=1.28 | Proposed feature and interaction testing method · Synthetic data | feature power: 0.999 -> 0.770 | Hurt |
| Increase model-output noise from σ=0.00 to σ=1.28 | Proposed feature and interaction testing method · Synthetic data | interaction power: 0.991 -> 0.564 | Hurt |
| Increase synthetic test-set size across the evaluated values | Proposed feature and interaction testing method · Synthetic data | feature FDR: 0.019 at m=32; 0.032 at m=16384 | No clear effect |
| Increase model-output noise across the evaluated values | Proposed feature and interaction testing method · Synthetic data | interaction FDR: 0.000 at σ=0.00; 0.050 at σ=1.28 | No clear effect |
| Test interactions only after constituent features are identified as important | Proposed interaction analysis · Synthetic data | power | Hurt |
Closest papers by meaning
- Adjustment for Confounding using Pre-Trained Representations (ICML 2025 · similarity 0.85)
- Permutation-Based Hypothesis Testing for Neural Networks (AAAI 2024 · similarity 0.85)
- On the Joint Interaction of Models, Data, and Features (ICLR 2024 · similarity 0.84)
- Additive Models Explained: A Computational Complexity Approach (NeurIPS 2025 · similarity 0.84)
- Exploring the cloud of feature interaction scores in a Rashomon set (ICLR 2024 · similarity 0.84)
- Conditional Feature Importance with Generative Modeling Using Adversarial Random Forests (AAAI 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
