{"about":{"site":"https://codewithpapers.app","non_affiliation":"Code with Papers and Syntology are not affiliated with, endorsed by, or sponsored by Papers with Code, Meta, or the pwc-archive mirror.","licence":"CC BY-SA 4.0","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","attribution":"https://codewithpapers.app/attribution","modified":"archive material modified by Syntology; see the attribution page"},"url":"/paper/reason-rft-reinforcement-fine-tuning-for","title":"Reason-RFT: Reinforcement Fine-Tuning for Visual Reasoning","arxiv_id":"2503.20752","date":"2025-03-26","proceeding":null,"authors":["Huajie Tan","Yuheng Ji","Xiaoshuai Hao","Minglan Lin","Pengwei Wang","Zhongyuan Wang","Shanghang Zhang"],"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 VLM 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 ability to transfer visual reasoning skills across domains and limiting its real-world applicability. To address these limitations, we propose Reason-RFT, a novel reinforcement fine-tuning framework that significantly enhances generalization capabilities in visual reasoning tasks. Reason-RFT introduces a two-phase training framework for visual reasoning: (1) Supervised Fine-Tuning (SFT) with curated Chain-of-Thought (CoT) data activates the reasoning potential of Vision-Language Models (VLMs), followed by (2) Group Relative Policy Optimization (GRPO)-based reinforcement learning that generates multiple reasoning-response pairs, significantly enhancing generalization in visual reasoning tasks. To evaluate Reason-RFT's visual reasoning capabilities, we reconstructed a comprehensive dataset spanning visual counting, structure perception, and spatial transformation. Experimental results demonstrate Reasoning-RFT's three key advantages: (1) Performance Enhancement: achieving state-of-the-art results across multiple tasks, outperforming most mainstream open-source and proprietary models; (2) Generalization Superiority: consistently maintaining robust performance across diverse tasks and domains, outperforming alternative training paradigms; (3) Data Efficiency: excelling in few-shot learning scenarios while surpassing full-dataset SFT baselines. Project website: https://tanhuajie.github.io/ReasonRFT","url_abs":"https://arxiv.org/abs/2503.20752v2","url_pdf":"https://arxiv.org/pdf/2503.20752v2.pdf","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","licence_url":"https://creativecommons.org/licenses/by-sa/4.0/legalcode","row_kind":"abstracts"},"code_links":[{"paper_slug":"reason-rft-reinforcement-fine-tuning-for","repo_url":"https://github.com/tanhuajie/reason-rft","is_official":0,"mentioned_in_paper":0,"mentioned_in_github":1,"framework":"none","reach":{"status":"ok","spdx":"Apache-2.0"}}],"tasks":[{"task_slug":"few-shot-learning","task_name":"Few-Shot Learning"},{"task_slug":"visual-reasoning","task_name":"Visual Reasoning"}],"methods":[{"method_slug":"sft","method_name":"SFT"}],"datasets_introduced":[],"methods_introduced":[],"results":[],"syntology":{"syntology_url":"https://syntology.ai/paper/2503.20752","atlas_url":"https://app.syntology.ai/?focus=2503.20752","mcp":{"get_harvested_code_for_paper":{"arxiv_id":"2503.20752"}},"developers":"https://syntology.ai/developers","read_at":"2026-09-25T09:33:49+00:00","read_at_is":"when the build read Syntology's graph, not when any sample ran","claim":"Per-sample execution status on synthesized fixtures; not a correctness claim about the paper. Samples come from repositories linked to the paper, official or community; repo_kind says which.","repos":[{"provenance":"external:paperswithcode_snapshot_2025-07-28","url":"https://github.com/tanhuajie/reason-rft","reach":{"status":"ok","spdx":"Apache-2.0"}}],"summary":{"ran":4,"unverified":4},"by_repo_kind":{"listed":{"samples":8,"ran":4,"repositories":1}},"repo_kind_vocabulary":{"official":"The archive marks this repository official for the paper","named_in_paper":"The archive records that the paper mentions this repository; it is not marked official","listed":"In the archive's code links for this paper, not marked official and not recorded as mentioned in the paper","found_in_text":"Syntology found this repository in the paper's own text; whether it is the authors' implementation is not asserted","community":"Not in the archive's code links for this paper; a community repository Syntology harvested"},"n_pointer_only_for_licence":0,"samples":[{"code_sha256_prefix":"b9da4075b3b8bdc4","entry":"extract_items","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"eval/cal_score_benchmarks_for_open_source.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/eval/cal_score_benchmarks_for_open_source.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":true,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"b9da4075b3b8bdc4"}},{"code_sha256_prefix":"e7f67f17660675fa","entry":"func_accuracy_reward","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"train/stage_rl/reward.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/train/stage_rl/reward.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"e7f67f17660675fa"}},{"code_sha256_prefix":"647e67d2dd82ab4b","entry":"get_image_data_url","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/distill_cot_data.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/distill_cot_data.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"647e67d2dd82ab4b"}},{"code_sha256_prefix":"bb5c5b6e897e6cd5","entry":"transform_item","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/convert_qwen2vl_format.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/convert_qwen2vl_format.py","link_basis":"harvester_set","language":"python","status":"ran","verification_level":1,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"bb5c5b6e897e6cd5"}},{"code_sha256_prefix":"2fbb95380e6fd336","entry":"gpt4o_query","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/distill_cot_data.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/distill_cot_data.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"2fbb95380e6fd336"}},{"code_sha256_prefix":"ef6e2e51fb8f04b5","entry":"gpt4o_query","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/distill_cot_data_trance.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/distill_cot_data_trance.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"ef6e2e51fb8f04b5"}},{"code_sha256_prefix":"4e7ac788ac1a18f4","entry":"process_item","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/distill_cot_data.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/distill_cot_data.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"4e7ac788ac1a18f4"}},{"code_sha256_prefix":"752bd3dc3a39db9a","entry":"process_item","repo":"tanhuajie/reason-rft","repo_kind":"listed","path":"utils/distill_cot_data_trance.py","file_url":"https://github.com/tanhuajie/reason-rft/blob/HEAD/utils/distill_cot_data_trance.py","link_basis":"harvester_set","language":"python","status":"unverified","verification_level":0,"contract_check":null,"metamorphic_tier":null,"behaviour_fingerprint":false,"licence":"Apache-2.0","inline_ok":true,"mcp_get_code":{"code_sha256":"752bd3dc3a39db9a"}}]},"arxiv_metadata":null,"syntology_extracted_results":null}