Papers
arxiv:2512.06835

Decouple to Generalize: Context-First Self-Evolving Learning for Data-Scarce Vision-Language Reasoning

Published on Dec 7
ยท Submitted by TINGYU LI on Dec 9
Authors:
,
,
,

Abstract

DoGe, a dual-decoupling framework, enhances vision-language models by separating context learning from problem solving, using a curriculum learning pipeline to improve reward signals and data diversity.

AI-generated summary

Recent vision-language models (VLMs) achieve remarkable reasoning through reinforcement learning (RL), which provides a feasible solution for realizing continuous self-evolving large vision-language models (LVLMs) in the era of experience. However, RL for VLMs requires abundant high-quality multimodal data, especially challenging in specialized domains like chemistry, earth sciences, and multimodal mathematics. Existing strategies such as synthetic data and self-rewarding mechanisms suffer from limited distributions and alignment difficulties, ultimately causing reward hacking: models exploit high-reward patterns, collapsing policy entropy and destabilizing training. We propose DoGe (Decouple to Generalize), a dual-decoupling framework that guides models to first learn from context rather than problem solving by refocusing on the problem context scenarios overlooked by synthetic data methods. By decoupling learning process into dual components (Thinker and Solver), we reasonably quantify the reward signals of this process and propose a two-stage RL post-training approach from freely exploring context to practically solving tasks. Second, to increase the diversity of training data, DoGe constructs an evolving curriculum learning pipeline: an expanded native domain knowledge corpus and an iteratively evolving seed problems pool. Experiments show that our method consistently outperforms the baseline across various benchmarks, providing a scalable pathway for realizing self-evolving LVLMs.

Community

Paper author Paper submitter
โ€ข
edited about 7 hours ago

abstract


Experiment Results ๐Ÿ“Š

We evaluate DoGe on 7 benchmarks covering:

  • General visual reasoning & hallucination (MMMU, MMStar, HallBench)
  • Specialized domain reasoning (MathVision, MathVista, ChemBench, MSEarthMCQ)

3B-level Models Performance

Method MMMU MMStar HallBench MathVision MathVista ChemBench MSEarthMCQ Avg.
InternVL2.5-2B 43.6 53.7 42.6 13.5 51.3 - - -
Visionary-3B 40.7 50.5 59.8 17.1 54.7 40.8 38.2 43.1
Qwen2.5VL-3B* (Base) 41.0 49.3 60.6 18.7 48.8 43.4 40.8 43.2
DoGe-3B (Iter1) 46.6 54.5 61.5 21.7 ๐Ÿฅ‡57.9 45.8 ๐Ÿฅ‡48.3 48.0
DoGe-3B (Iter2) 48.9 52.5 ๐Ÿฅ‡62.5 23.1 54.2 ๐Ÿฅ‡47.7 46.2 47.9
DoGe-3B (Iter3) ๐Ÿฅ‡50.2 ๐Ÿฅ‡54.7 61.8 ๐Ÿฅ‡24.2 57.0 46.9 47.3 ๐Ÿฅ‡48.9
โฌ†๏ธ Max Gain (vs. Base) +9.2 +5.4 +1.9 +5.5 +9.1 +4.3 +7.5 +5.7

7B-level Models Performance

Method MMMU MMStar HallBench MathVision MathVista ChemBench MSEarthMCQ Avg.
InternVL2.5-8B 48.9 62.8 50.1 22.0 64.4 - - -
Vision-R1-7B 46.9 60.8 66.7 ๐Ÿฅ‡29.0 68.5 46.0 44.1 51.7
Qwen2.5VL-7B* (Base) 49.9 60.7 66.3 23.6 64.1 48.6 43.3 50.9
DoGe-7B (Iter1) 53.1 ๐Ÿฅ‡63.2 54.4 24.3 62.1 48.7 46.4 50.3
DoGe-7B (Iter2) 50.9 60.0 ๐Ÿฅ‡68.3 25.3 ๐Ÿฅ‡68.8 ๐Ÿฅ‡49.0 ๐Ÿฅ‡46.5 52.7
DoGe-7B (Iter3) ๐Ÿฅ‡53.6 63.0 68.0 25.2 68.3 48.5 45.8 ๐Ÿฅ‡53.2
โฌ†๏ธ Max Gain (vs. Base) +3.7 +2.5 +2.0 +1.7 +4.7 +0.4 +3.2 +2.3

Key Takeaways โœจ

  1. Stable Self-Evolution: DoGe achieves consistent performance improvement across 3 iterations for both 3B and 7B models
  2. Domain Generalization:
    • 3B models: Average +5.7% performance gain across all benchmarks
    • 7B models: Average +2.3% performance gain (maintains superiority over strong baselines)
  3. Hallucination Reduction: +2.0% average improvement on HallBench, mitigating visual hallucination
  4. Data Efficiency: Excels in data-scarce domains (Chemistry, Earth Science) with limited manual annotations

Sign up or log in to comment

Models citing this paper 0

No model linking this paper

Cite arxiv.org/abs/2512.06835 in a model README.md to link it from this page.

Datasets citing this paper 0

No dataset linking this paper

Cite arxiv.org/abs/2512.06835 in a dataset README.md to link it from this page.

Spaces citing this paper 0

No Space linking this paper

Cite arxiv.org/abs/2512.06835 in a Space README.md to link it from this page.

Collections including this paper 0

No Collection including this paper

Add this paper to a collection to link it from this page.