PROPA: Toward Process-level Optimization in Visual Reasoning via Reinforcement Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Jiang, Yanbei, Lei, Chao, Ding, Yihao, Ehinger, Krista, Lau, Jey Han
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866911263695568896
author Jiang, Yanbei
Lei, Chao
Ding, Yihao
Ehinger, Krista
Lau, Jey Han
author_facet Jiang, Yanbei
Lei, Chao
Ding, Yihao
Ehinger, Krista
Lau, Jey Han
contents Despite significant progress, Vision-Language Models (VLMs) still struggle with complex visual reasoning, where multi-step dependencies cause early errors to cascade through the reasoning chain. Existing post-training paradigms are limited: Supervised Fine-Tuning (SFT) relies on costly step-level annotations, while Reinforcement Learning with Verifiable Rewards (RLVR) methods like GRPO provide only sparse, outcome-level feedback, hindering stable optimization. We introduce PROPA (Process-level Reasoning Optimization with interleaved Policy Alignment), a novel framework that integrates Monte Carlo Tree Search (MCTS) with GRPO to generate dense, process-level rewards and optimize reasoning at each intermediate step without human annotations. To overcome the cold-start problem, PROPA interleaves GRPO updates with SFT, enabling the model to learn from both successful and failed reasoning trajectories. A Process Reward Model (PRM) is further trained to guide inference-time search, aligning the test-time search with the training signal. Across seven benchmarks and four VLM backbones, PROPA consistently outperforms both SFT- and RLVR-based baselines. It achieves up to 17.0% gains on in-domain tasks and 21.0% gains on out-of-domain tasks compared to existing state-of-the-art, establishing a strong reasoning and generalization capability for visual reasoning tasks. The code isavailable at: https://github.com/YanbeiJiang/PROPA.
format Preprint
id arxiv_https___arxiv_org_abs_2511_10279
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle PROPA: Toward Process-level Optimization in Visual Reasoning via Reinforcement Learning
Jiang, Yanbei
Lei, Chao
Ding, Yihao
Ehinger, Krista
Lau, Jey Han
Computer Vision and Pattern Recognition
Despite significant progress, Vision-Language Models (VLMs) still struggle with complex visual reasoning, where multi-step dependencies cause early errors to cascade through the reasoning chain. Existing post-training paradigms are limited: Supervised Fine-Tuning (SFT) relies on costly step-level annotations, while Reinforcement Learning with Verifiable Rewards (RLVR) methods like GRPO provide only sparse, outcome-level feedback, hindering stable optimization. We introduce PROPA (Process-level Reasoning Optimization with interleaved Policy Alignment), a novel framework that integrates Monte Carlo Tree Search (MCTS) with GRPO to generate dense, process-level rewards and optimize reasoning at each intermediate step without human annotations. To overcome the cold-start problem, PROPA interleaves GRPO updates with SFT, enabling the model to learn from both successful and failed reasoning trajectories. A Process Reward Model (PRM) is further trained to guide inference-time search, aligning the test-time search with the training signal. Across seven benchmarks and four VLM backbones, PROPA consistently outperforms both SFT- and RLVR-based baselines. It achieves up to 17.0% gains on in-domain tasks and 21.0% gains on out-of-domain tasks compared to existing state-of-the-art, establishing a strong reasoning and generalization capability for visual reasoning tasks. The code isavailable at: https://github.com/YanbeiJiang/PROPA.
title PROPA: Toward Process-level Optimization in Visual Reasoning via Reinforcement Learning
topic Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2511.10279