AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

AQuA: Toward Strategic Response Generation for Ambiguous Visual Questions

Jihyoung Jang, Hyounghun Kim

ICLR 2026 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 0 citations (OpenAlex)

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 changedWhereMeasuredEffect
add GRPO after SFT (ablation)Qwen2.5-VL-3B-Tuned · AQuA testoverall strategic accuracy: 83.81 -> 86.28Improved
add GRPO after SFT (ablation)InternVL3-2B-Tuned · AQuA testoverall strategic accuracy: 73.42 -> 79.11Improved
add GRPO after SFT (ablation)Qwen2.5-VL-3B-Tuned · AQuA test (Level 2)strategic accuracy: 61.33 -> 82.22Improved
add GRPO after SFT (ablation)InternVL3-2B-Tuned · AQuA test (Level 2)strategic accuracy: 37.67 -> 59.67Improved
add GRPO after SFT (ablation)Qwen2.5-VL-3B-Tuned · AQuA test (Level 1)strategic accuracy: 92.22 -> 77.0Hurt
add Chain-of-Thought promptQwen2.5-VL-3B-Instruct · AQuA testoverall strategic accuracy: 32.83 -> 28.42Hurt
add Chain-of-Thought promptGPT-5 · AQuA testoverall strategic accuracy: 22.86 -> 25.72No clear effect
add strategy promptingQwen2.5-VL-3B-Instruct · AQuA testoverall strategic accuracy: 32.83 -> 25.42No clear effect
add strategy promptingGPT-5 · AQuA testoverall strategic accuracy: 22.86 -> 42.25Improved
add strategy promptingQwen2.5-VL-72B-Instruct · AQuA testoverall strategic accuracy: 25.78 -> 42.22Improved
fine-tune on AQuA (SFT+GRPO) instead of zero-shotQwen2.5-VL-3B-Instruct · AQuA testoverall strategic accuracy: 32.83 -> 86.28Improved
fine-tune on AQuA (SFT+GRPO) instead of zero-shotInternVL3-2B-Instruct · AQuA testoverall strategic accuracy: 25.95 -> 79.11Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy