홈 › Papers

Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning

2026-06-30 · Junha Jung, Minbyul Jeong, Suhyeon Lim, Sungwook Jung, Jaehoon Yun, Taeyun Roh, Mujeen Sung, Jaewoo Kang hf

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 three multimodal LLM backbones, MRPO consistently outperforms standard GRPO and a recent RL baseline, and on Qwen3-VL-8B-Instruct even surpasses substantially larger medical MLLMs such as HuatuoGPT-Vision-34B by 2.79 points. Moreover, MRPO reduces early-stage reasoning failures from 64.0% to 13.0%, showing that targeted mitigation of cascading failures improves both reasoning quality and final answer accuracy. Our code is available at https://github.com/dmis-lab/MRPO

📄 PDF Abstract BibTeX arXiv:2606.31825

Code (0)

등록된 구현이 없습니다.

Tasks

Visual Question Answering Reinforcement Learning Multimodal Reasoning

Graph Cascades: Contagion-Based Mesoscopic Rewiring for Structure-Aware Graph Machine Learning

2026-06-03 · Meher Chaitanya, My Le, Luana Ruiz arxiv

We introduce Graph Cascades, a mesoscopic rewiring strategy for Graph Neural Networks (GNNs) and Graph Transformers (GTs) that captures intermediate-scale graph structure beyond purely local edges or fully global attenti…

Failure-Aware RL: Reliable Offline-to-Online Reinforcement Learning with Self-Recovery for Real-World Manipulation

2026-01-12 · Huanyu Li, Kun Lei, Sheng Zang, Kaizhe Hu 외 arxiv

Post-training algorithms based on deep reinforcement learning can push the limits of robotic models for specific objectives, such as generalizability, accuracy, and robustness. However, Intervention-requiring Failures (I…

Reinforcement Learning

AgentForesight: Online Auditing for Early Failure Prediction in Multi-Agent Systems

2026-05-09 · Boxuan Zhang, Jianing Zhu, Zeru Shi, Dongfang Liu 외 arxiv

LLM-based multi-agent systems are increasingly deployed on long-horizon tasks, but a single decisive error is often accepted by downstream agents and cascades into trajectory-level failure. Existing work frames this as \…

Reinforcement Learning

Instance-aware Semantic Segmentation via Multi-task Network Cascades

2015-12-14 · CVPR 2016 6 · Jifeng Dai, Kaiming He, Jian Sun

Semantic segmentation research has recently witnessed rapid progress, but many leading methods are unable to identify object instances. In this paper, we present Multi-task Network Cascades for instance-aware semantic se…

Multi-Human Parsing object-detection Object Detection Segmentation+1

Power Failure Cascade Prediction using Graph Neural Networks

2024-04-24 · Sathwik Chadaga, Xinyu Wu, Eytan Modiano

We consider the problem of predicting power failure cascades due to branch failures. We propose a flow-free model based on graph neural networks that predicts grid states at every generation of a cascade process given an…

Graph Neural Network Prediction