AI Paper Atlas

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

Beyond Multiple Choice: Verifiable OpenQA for Robust Vision-Language RFT

Yesheng Liu, Hao Li, Haiyu Xu, Baoqi Pei, Jiahao Wang, Mingxuan Zhao, Jing-Shu Zheng, Zheqi He, JG Yao, Xi Yang, Bowen Qin, Jiajun Zhang

CVPR 2026 · Vision-language and multimodal › Visual question answering · approach: Universal VQA models · 1 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
RFT on MCQA data instead of base model (ablation)Qwen2.5-VL-7B · MMMUMCQA-OpenQA gap: 30.2 -> 39.3 gap; open 21.4 -> 17.1Hurt
RFT on MCQA data instead of base model (ablation)Qwen2.5-VL-3B · MMLU-ProMCQA-OpenQA gap: 18.4 -> 27.0Hurt
train on ReVeL-rewritten OpenQA instead of MCQA (ablation)Qwen2.5-VL-7B · Overall open-ended accuracyOpenQA accuracy: 24.7 -> 34.0Improved
train on ReVeL-rewritten OpenQA instead of MCQA (ablation)Qwen2.5-VL-7B · Overall MCQA accuracyMCQA accuracy: 47.8 -> 46.8No clear effect
train on ReVeL-rewritten OpenQA instead of MCQA (ablation)Qwen2.5-VL-3B · Overall scoreoverall score: 30.1 -> 34.3Improved
train on ReVeL-rewritten OpenQA instead of original ViRL OpenQA (ablation)Qwen2.5-VL-3B · Overall open-ended accuracyOpenQA accuracy: 21.8 -> 28.0Improved
add original ViRL OpenQA data on top of ReVeL OpenQA (ablation)Qwen2.5-VL-7B · Overall scoreoverall score: 40.4 -> 39.8No clear effect
train InternVL3-4B on ReVeL OpenQAInternVL3-4B · Overall scoreoverall score: 42.4 -> 42.8No clear effect
use rule-based hybrid judging instead of pure LLM judgeReVeL evaluation · 600 sampled responsesjudge accuracy: 97.3 -> 98.5Improved
add answer options to open-ended questionsfrontier LLMs and VLMs · SimpleQA, VisualSimpleQAaccuracyImproved
replace ground-truth option with 'None of the above'VLMs · MMLU-Pro, MMMUreasoning-answer mismatch rate: 18% -> 50%Hurt
switch evaluation from MCQA to OpenQAGPT-5 · MMMUaccuracy: 79.2 -> 59.5Hurt

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.

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

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