大模型
arxiv.org ·
打破失败级联：面向医学多模态推理的步骤感知强化学习
该论文指出医学视觉问答中早期推理错误会级联导致最终答案错误，而现有后训练流程以结果为中心、信用分配稀疏。作者提出 MRPO（Medical Reasoning-aware Policy Optimization），在最终答案错误时对更早的无效推理步骤 token 施加指数级更大惩罚，从而打断失败级联且不损害成功路径。摘要称在四个多模态 LLM 骨干上持续优于标准 GRPO 和近期 RL 基线，Qwen3-VL-8B-Thinking 在医学 VQA 上超过 HuatuoGPT-Vision-34B 4.59 分，早期推理失败率由 58.6% 降至 13.4%。代码已公开。
AI推荐理由 arXiv 论文（v2，2026-09-24 修订），作者与摘要明确，属可验证的研究首发。核心新信息是 MRPO：在最终答案错误时对早期无效推理步骤的 token 施加指数级更大惩罚，以打断医学多模态推理中的失败级联；摘要称在四个多模态 LLM 骨干上优于标准 GRPO 与一个近期 RL 基线，Qwen3-VL-8B-Thinking 在医学 VQA 上超过 HuatuoGPT-Vision-34B 4.59 分，并将早期推理失败率从 58.6% 降至 13.4%。局限：这些数字均来自论文自述，未提供独立复现或基准细节，且仅限医学 VQA 场景，通用性待验证。
78
正文
Computer Science > Computer Vision and Pattern Recognition
arXiv:2606.31825 (cs)
[Submitted on 30 Jun 2026 (v1), last revised 24 Sep 2026 (this version, v2)]
Title:Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning
Authors:Junha Jung, Minbyul Jeong, Suhyeon Lim, Sungwook Jung, Jaehoon Yun, Taeyun Roh, Mujeen Sung, Jaewoo Kang
View a PDF of the paper titled Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning, by Junha Jung and 7 other authors
View PDF
HTML (experimental)
Abstract:Recent multimodal large language models have shown great promise in clinical image reasoning, but existing post-training pipelines remain predominantly outcome-centric, relying on final answer correctness or sequence-level preferences. This suffers from sparse credit assignment, making it difficult to optimize the reasoning process essential for clinical applications. Our analysis reveals that cascading errors from early-stage reasoning failures are a leading cause of incorrect predictions in medical visual question answering (VQA) benchmarks. Motivated by this, we propose Medical Reasoning-aware Policy Optimization (MRPO), an RL algorithm that incorporates step-wise process rewards. When the final answer is incorrect, MRPO assigns exponentially larger penalties to tokens in earlier invalid reasoning steps, breaking failure cascades without compromising successful paths. Across four multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Thinking even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 4.59 points. Moreover, MRPO reduces early-stage reasoning failures from 58.6% to 13.4%, showing that targeted mitigation of cascading failures improves both reasoning quality and final answer accuracy. Our code is available at this https URL
Subjects:
Computer Vision and Pattern Recognition (cs.CV); Artificial Intelligence (cs.AI)
Cite as:
arXiv:2606.31825 [cs.CV]
(or
arXiv:2606.31825v2 [cs.CV] for this version)
https://doi.org/10.48550/arXiv.2606.31825
Focus to learn more
arXiv-issued DOI via DataCite
Submission history
From: Junha Jung [view email]
[v1]
Tue, 30 Jun 2026 15:35:19 UTC (2,979 KB)
[v2]
Thu, 24 Sep 2026 16:20:32 UTC (3,476 KB)
Full-text links:
Access Paper:
View a PDF of the paper titled Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning, by Junha Jung and 7 other authors
View PDF
HTML (experimental)
TeX Source
view license
Current browse context:
cs.CV
new
|
recent
| 2026-06
Change to browse by:
cs
cs.AI
References & Citations
NASA ADS
Google Scholar
Semantic Scholar
export BibTeX citation
Loading...
BibTeX formatted citation
×
loading...
Data provided by:
Bookmark
Bibliographic Tools
Bibliographic and Citation Tools
Bibliographic Explorer Toggle
Bibliographic Explorer (What is the Explorer?)
Connected Papers Toggle
Connected Papers (What is Connected Papers?)
Litmaps Toggle
Litmaps (What is Litmaps?)
scite.ai Toggle
scite Smart Citations (What are Smart Citations?)
Code, Data, Media
Code, Data and Media Associated with this Article
alphaXiv Toggle
alphaXiv (What is alphaXiv?)
Links to Code Toggle
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub Toggle
DagsHub (What is DagsHub?)
GotitPub Toggle
Gotit.pub (What is GotitPub?)
Huggingface Toggle
Hugging Face (What is Huggingface?)
ScienceCast Toggle
ScienceCast (What is ScienceCast?)
Demos
Demos
Replicate Toggle
Replicate (What is Replicate?)
Spaces Toggle
Hugging Face Spaces (What is Spaces?)
Spaces Toggle
TXYZ.AI (What is TXYZ.AI?)
Related Papers
Recommenders and Search Tools
Link to Influence Flower
Influence Flower (What are Influence Flowers?)
Core recommender toggle
CORE Recommender (What is CORE?)
Author
Venue
Institution
Topic
About arXivLabs
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Which authors of this paper are endorsers? |
Disable MathJax (What is MathJax?)