AI Paper Atlas

Research map / Reinforcement learning / RL from feedback

DIVA-GRPO: Enhancing Multimodal Reasoning through Difficulty-Adaptive Variant Advantage

Haowen Gao, zhenyu zhang, Liang Pang, Fangda Guo, douhongjian, Guannan Lv, ShaoGuo Liu, Tingting Gao, Huawei Shen, Xueqi Cheng

ICLR 2026 · Reinforcement learning › RL from feedback · approach: GRPO stability and variants · 3 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 adaptive variant generation (ablation)DIVA-GRPO · Ablation average (MathVista/MathVerse/MMK12test)accuracy (%): 61.6 -> 66.1Improved
add difficulty-weighted scaling (ablation)DIVA-GRPO · Ablation averageaccuracy (%): 64.0 -> 66.1Improved
add RRB-Rescaling (ablation)DIVA-GRPO · Ablation averageaccuracy (%): 63.8 -> 66.1Improved
add global-local balance (ablation)DIVA-GRPO · Ablation averageaccuracy (%): 64.1 -> 66.1Improved
add RRB-Rescaling to standard GRPOGRPO · MathVista/MathVerse/MMK12test averageaccuracy (%): 60.23 -> 62.17Improved
train with DIVA-GRPO instead of GRPOQwen3-VL-4B-Instruct · Math/Physics/Chemistry/Biology averageaccuracy (%): 70.9 -> 73.4Improved
train with DIVA-GRPO instead of GRPOQwen3-VL-8B-Instruct · Math/Physics/Chemistry/Biology averageaccuracy (%): 74.3 -> 78.5Improved
fine-tune with DIVA-GRPOQwen-2.5-VL-7B · six multimodal math benchmarks averageaccuracy (%): 46.23 -> 54.58Improved
direct SFT on GPT-o3 reasoning tracesQwen-2.5-VL-7B · six multimodal math benchmarks averageaccuracy (%): 46.23 -> 44.23Hurt
generate variants with GPT-o3 instead of Qwen-Plus (ablation)DIVA-GRPO · Math/Physics/Chemistry/Biology averageaccuracy (%): 69.8 -> 70.2No clear effect
use relative instead of absolute difficulty weighting (ablation)DIVA-GRPO · validationaccuracyNo clear effect
set sensitivity k to 0.1 rather than 0 or 1 (ablation)DIVA-GRPO · validationaccuracyMixed

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