Research map / Vision-language and multimodal / Visual question answering
Self-Critical Reasoning for Robust Visual Question Answering
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 |
|---|---|---|---|
| add influence-strengthening loss only (ablation) | UpDn · VQA-CP v2 test | accuracy (All): 39.7 -> 47.8 | Helped |
| add self-critical loss on top of influence strengthening (ablation) | UpDn+SCR · VQA-CP v2 test | accuracy (All): 47.8 -> 49.5 | Helped |
| increase lambda_crit 2000 -> 6000 (ablation) | UpDn+SCR · VQA-CP v2 test | accuracy (All): 49.5 -> 49.0 | No clear effect |
| vary proposal set size 4-10 (ablation) | UpDn+SCR (HAT) · VQA-CP v2 test | accuracy (All): 48.3%-49.2% | No clear effect |
| add self-critical training with VQA-X explanations | UpDn · VQA-CP v2 test | accuracy (Num): 11.9 -> 10.93 | Hurt |
| add self-critical training | UpDn · VQA v2 val | accuracy (All): 63.5 -> 62.2 | Hurt |
| add self-critical training | UpDn · VQA-CP v2 test | false sensitivity rate: 35.5% -> 19.6% | Helped |
| use SCR instead of HINT with same HAT explanations | UpDn · VQA-CP v2 test | accuracy (All): 47.7 -> 49.17 | Helped |
| use textual VQA-X instead of visual HAT explanations | UpDn+SCR · VQA-CP v2 test | accuracy (All): 49.17 -> 49.45 | Helped |
Closest papers by meaning
- Don't Just Assume; Look and Answer: Overcoming Priors for Visual Question Answering (CVPR 2018 · similarity 0.93)
- Tips and Tricks for Visual Question Answering: Learnings From the 2017 Challenge (CVPR 2018 · similarity 0.93)
- Answer Them All! Toward Universal Visual Question Answering Models (CVPR 2019 · similarity 0.93)
- Visual Question Answering with Question Representation Update (QRU) (NeurIPS 2016 · similarity 0.93)
- Answer-Type Prediction for Visual Question Answering (CVPR 2016 · similarity 0.93)
- When Open-Vocabulary Visual Question Answering Meets Causal Adapter: Benchmark and Approach (AAAI 2025 · similarity 0.92)
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.