SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2503.20752 · NeurIPS · 2025

Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning of Vision Language Models

Shanghang Zhang, Zhongyuan Wang, Xiaoshuai Hao, Huajie Tan, Yuheng Ji, Xiansheng Chen, Pengwei Wang, Sam Sam

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 8 functions out of this paper's own repositories and ran 4 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
tanhuajie/reason-rft pwc_unofficial 4 of 8
FunctionStatusWhere it lives
extract_items Ran tanhuajie/reason-rft/eval/cal_score_benchmarks_for_open_source.py
code served (permissive licence) · get_code("b9da4075b3b8bdc4")
func_accuracy_reward Ran tanhuajie/reason-rft/train/stage_rl/reward.py
code served (permissive licence) · get_code("e7f67f17660675fa")
get_image_data_url Ran tanhuajie/reason-rft/utils/distill_cot_data.py
code served (permissive licence) · get_code("647e67d2dd82ab4b")
transform_item Ran tanhuajie/reason-rft/utils/convert_qwen2vl_format.py
code served (permissive licence) · get_code("bb5c5b6e897e6cd5")
gpt4o_query Not yet run tanhuajie/reason-rft/utils/distill_cot_data.py
code served (permissive licence) · get_code("2fbb95380e6fd336")
gpt4o_query Not yet run tanhuajie/reason-rft/utils/distill_cot_data_trance.py
code served (permissive licence) · get_code("ef6e2e51fb8f04b5")
process_item Not yet run tanhuajie/reason-rft/utils/distill_cot_data.py
code served (permissive licence) · get_code("4e7ac788ac1a18f4")
process_item Not yet run tanhuajie/reason-rft/utils/distill_cot_data_trance.py
code served (permissive licence) · get_code("752bd3dc3a39db9a")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Visual reasoning abilities play a crucial role in understanding complex multimodal data, advancing both domain-specific applications and artificial general intelligence (AGI). Existing methods improve Vision-Language Models (VLMs) reasoning via Chain-of-Thought (CoT) supervised fine-tuning, using meticulously annotated training data to enhance visual reasoning capabilities. However, this training paradigm may lead to overfitting and cognitive rigidity, restricting the model's generalization ability to transfer visual reasoning skills under domain shift and limiting its real-world applicability. To address these limitations, we propose Reason-RFT, the first two-stage reinforcement fine-tuning framework for visual reasoning: (1) Supervised Fine-Tuning (SFT) with curated CoT data activates the reasoning potential of VLMs, followed by (2) Group Relative Policy Optimization (GRPO)-based reinforcement learning that generates multiple reasoning-response pairs, significantly enhancing the capability to address ubiquitous domain shift in visual reasoning tasks. To evaluate the visual reasoning capabilities of Reason-RFT, we reconstructed a comprehensive dataset encompassing visual counting, structural perception, and spatial transformation, serving as a benchmark for systematic assessment across three core dimensions. Experimental results demonstrate three key advantages: (1) Performance Enhancement: achieving state-of-the-art results across multiple tasks, outperforming mainstream open-source and proprietary models; (2) Generalization Superiority: consistently maintaining robust performance in addressing domain shift in typical visual reasoning tasks, outperforming alternative paradigms; (3) Data Efficiency: excelling in few-shot learning scenarios while surpassing full-dataset SFT baselines. Reason-RFT introduces a rebust training paradigm in visual reasoning, and please refer to project website: Reason-RFT.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2503.20752")
get_code_for_paper("2503.20752")
have("2503.20752")

Connect an agent — have() is free.