AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

From Blind Spots to Gains: Diagnostic-Driven Iterative Training for Large Multimodal Models

Hongrui Jia, Chaoya Jiang, Yongrui Heng, Shikun Zhang, Wei Ye

ICML 2026 · Reinforcement learning › RL from feedback · approach: RL for LLM reasoning · 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 diagnostic module (ablation)DPE · CharXiv RQ (Qwen2.5-VL-7B)accuracy (%): 36.7 -> 40.91Improved
add diagnostic module (ablation)DPE · MathVision (Qwen2.5-VL-7B)accuracy (%): 25.99 -> 26.51Improved
add image retrieval and editing tools (ablation)DPE · CharXiv RQ (Qwen2.5-VL-7B)accuracy (%): 38.1 -> 40.91Improved
add image retrieval and editing tools (ablation)DPE · MathVision (Qwen2.5-VL-7B)accuracy (%): 26.18 -> 26.51Improved
use DPE diagnosis-driven 3K samples instead of static Vision-SR1 47K dataQwen2.5-VL-7B-Instruct · MMMUaccuracy (%): 54.8 -> 56.44Improved
use DPE diagnosis-driven 3K samples instead of static Vision-SR1 47K dataQwen2.5-VL-7B-Instruct · MathVistaaccuracy (%): 68.8 -> 69.5Improved
use DPE instead of VisPlay (Iter 3)Qwen2.5-VL-7B-Instruct · Average over 11 benchmarksaccuracy (%): 58.33 -> 59.29Improved
apply DPE to stronger backboneQwen3-VL-8B-Instruct · Average over 11 benchmarksaccuracy (%): 65.64 -> 68.04Improved
apply DPE to stronger backboneQwen3-VL-8B-Instruct · ChartQAaccuracy (%): 85.08 -> 84.80No clear effect
use DPE instead of VisPlay for question text diversity (Iter 3)question generation · text diversitymean pairwise cosine distance: 0.797 -> 0.850Improved
use DPE instead of VisPlay for question quality (Iter 3)question generation · LLM-judged quality score QSscore (1-5): 3.32 -> 4.80Improved

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