Research map / Label-efficient and robust learning / Recognition and long-tail
Grounding Visual Explanations
Lisa Anne Hendricks, Ronghang Hu, Trevor Darrell, Zeynep Akata
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 |
|---|---|---|---|
| add phrase-critic reranking to explanation sampling | explanation model · CUB human evaluation | % correct noun phrases: 76.64 -> 77.96 | Improved |
| add phrase-critic reranking to explanation sampling | explanation model · CUB human evaluation | % correct sentences: 52.10 -> 61.97 | Improved |
| use phrase-critic instead of grounding-model score averaging | grounding model · CUB human evaluation | % correct noun phrases: 76.32 -> 77.96 | Improved |
| use phrase-critic instead of grounding-model score averaging | grounding model · CUB human evaluation | % correct sentences: 49.85 -> 61.97 | Improved |
| use grounding model averaging instead of fluency-only baseline | explanation model · CUB human evaluation | % correct sentences: 52.10 -> 49.85 | Hurt |
| show bounding boxes with explanations (ablation) | Phrase Critic · CUB MTurk informativeness | share more informative: 41.9% more informative, 49.3% equally informative | Improved |
| add phrase-critic on top of grounding model | grounding model · CUB head region | Euclidean distance (lower better): 46.31 -> 20.26 | Improved |
| add phrase-critic on top of grounding model | grounding model · CUB eye region | % accuracy: 60.78 -> 56.72 | Hurt |
| add phrase-critic on top of grounding model | grounding model · CUB head region | % accuracy: 60.60 -> 74.06 | Improved |
| use phrase-critic instead of grounding model | grounding model · FOIL classification | accuracy: 56.68 -> 87.00 | Improved |
| use phrase-critic instead of grounding model | grounding model · FOIL word detection | accuracy: 39.80 -> 73.72 | Improved |
| use phrase-critic instead of grounding model | grounding model · FOIL word correction | accuracy: 8.80 -> 49.60 | Improved |
Closest papers by meaning
- Improved Visual Grounding through Self-Consistent Explanations (CVPR 2024 · similarity 0.86)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.86)
- Counterfactual Explanations for Misclassified Images: How Human and Machine Explanations Differ (Abstract Reprint) (AAAI 2024 · similarity 0.85)
- Rex-Thinker: Grounded Object Referring via Chain-of-Thought Reasoning (ICLR 2026 · similarity 0.84)
- Language Model as Visual Explainer (NeurIPS 2024 · similarity 0.84)
- A General Search-Based Framework for Generating Textual Counterfactual Explanations (AAAI 2024 · 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.
