Research map / Multi-agent systems and games / Human-AI and goal-conditioned
Explain Your Move: Understanding Agent Actions Using Specific and Relevant Feature Attribution
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 SARFA saliency maps rather than no saliency | Human chess-puzzle solving · Chess puzzles | Accuracy: 56.67% -> 72.41% | Helped |
| Use SARFA saliency maps rather than no saliency | Human chess-puzzle solving · Chess puzzles | Average time taken: 77.53 sec -> 67.02 sec | Helped |
| Use SARFA saliency maps rather than Greydanus et al. saliency maps | Human chess-puzzle solving · Chess puzzles | Accuracy: 40.84% -> 72.41% | Helped |
| Use SARFA saliency maps rather than Greydanus et al. saliency maps | Human chess-puzzle solving · Chess puzzles | Average time taken: 70.95 sec -> 67.02 sec | Helped |
| Use SARFA saliency maps rather than Iyer et al. saliency maps | Human chess-puzzle solving · Chess puzzles | Accuracy: 24.60% -> 72.41% | Helped |
| Use SARFA saliency maps rather than Iyer et al. saliency maps | Human chess-puzzle solving · Chess puzzles | Average time taken: 102.26 sec -> 67.02 sec | Helped |
| Remove a random human-expert-labeled non-salient piece from the position | SARFA · Chess saliency dataset | AUC: 0.92 -> 0.92 | No clear effect |
| Remove a random SARFA-labeled non-salient piece from the position | SARFA · Chess saliency dataset | AUC: 0.92 -> 0.92 | No clear effect |
| Combine specificity and relevance with the harmonic mean rather than using either component alone (ablation) | SARFA · Chess saliency dataset | Saliency-map accuracy | Helped |
| Combine specificity and relevance with the harmonic mean rather than the average (ablation) | SARFA · Chess saliency dataset | Saliency-map accuracy | Helped |
| Combine specificity and relevance with the harmonic mean rather than the geometric mean (ablation) | SARFA · Chess saliency dataset | Saliency-map accuracy | Helped |
| Combine specificity and relevance with the harmonic mean rather than the minimum (ablation) | SARFA · Chess saliency dataset | Saliency-map accuracy | Helped |
Closest papers by meaning
- Exploratory Not Explanatory: Counterfactual Analysis of Saliency Maps for Deep Reinforcement Learning (ICLR 2020 · similarity 0.90)
- Attention Trajectories as a Diagnostic Axis for Deep Reinforcement Learning (NeurIPS 2026 · similarity 0.89)
- Explaining Reinforcement Learning Agents through Counterfactual Action Outcomes (AAAI 2024 · similarity 0.83)
- Explainable Saliency: Articulating Reasoning with Contextual Prioritization (CVPR 2025 · similarity 0.83)
- DiffCAM: Data-Driven Saliency Maps by Capturing Feature Differences (CVPR 2025 · similarity 0.82)
- Saliency-R1: Enforcing Interpretable and Faithful Vision-language Reasoning via Saliency-map Alignment Reward (CVPR 2026 · similarity 0.82)
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.