Research map / Learning theory and algorithms / Prediction and decision losses
Trade-off Between Efficiency and Consistency for Removal-based Explanations
Yifan Zhang, Haowei He, Zhiquan Tan, Yang Yuan
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 |
|---|---|---|---|
| Use Harmonica 3rd instead of the strongest listed baseline | Removal-based explanations · IMDb | C^3 truthful gap: 1.390 -> 0.174 | Improved |
| Use Harmonica 3rd instead of the strongest listed baseline | Removal-based explanations · ImageNet | C^3 truthful gap: 0.430 -> 0.246 | Improved |
| Use Harmonica-anchor with k=9 instead of Integrated Gradients | ImageNet explanations · ImageNet | L2 interpretation error: 31.8 times lower | Improved |
| Increase the number of anchors (ablation) | Harmonica-anchor · IMDb, ImageNet, and SST-2 | L2 interpretation error | Improved |
| Fit separate local models across neighborhood subspaces (ablation) | Harmonica-anchor · IMDb, ImageNet, and SST-2 | L2 interpretation error | Improved |
| Restrict Harmonica-local to a local neighborhood (ablation) | Harmonica-local · Local neighborhood and outside neighborhood | Interpretation error | Mixed |
| Increase the anchor-consistency penalty (ablation) | Harmonica-anchor-constrained · Removal-based explanations | Interpretation error | Hurt |
| Use Harmonica 3rd rather than Harmonica 2nd (ablation) | Harmonica · IMDb | C^3 truthful gap: 0.540 -> 0.174 | Improved |
| Use Harmonica 3rd rather than Harmonica 2nd (ablation) | Harmonica · ImageNet | C^3 truthful gap: 0.660 -> 0.246 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Faith-SHAP — Faith-Shap: The Faithful Shapley Interaction Index (ICML 2023) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A New Approach to Backtracking Counterfactual Explanations: A Unified Causal Framework for Efficient Model Interpretability (ICML 2025 · similarity 0.85)
- BEE: Metric-Adapted Explanations via Baseline Exploration-Exploitation (AAAI 2025 · similarity 0.84)
- Everything, Everywhere, All at Once: Is Mechanistic Interpretability Identifiable? (ICLR 2025 · similarity 0.84)
- Delivering Inflated Explanations (AAAI 2024 · similarity 0.84)
- Local vs. Global Interpretability: A Computational Complexity Perspective (ICML 2024 · similarity 0.84)
- A Variable Occurrence-Centric Framework for Inconsistency Handling (AAAI 2025 · similarity 0.83)
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.
