SPARK: Synergistic Policy And Reward Co-Evolving Framework

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Ziyu, Zang, Yuhang, Ding, Shengyuan, Cao, Yuhang, Dong, Xiaoyi, Duan, Haodong, Lin, Dahua, Wang, Jiaqi
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908561710252032
author Liu, Ziyu
Zang, Yuhang
Ding, Shengyuan
Cao, Yuhang
Dong, Xiaoyi
Duan, Haodong
Lin, Dahua
Wang, Jiaqi
author_facet Liu, Ziyu
Zang, Yuhang
Ding, Shengyuan
Cao, Yuhang
Dong, Xiaoyi
Duan, Haodong
Lin, Dahua
Wang, Jiaqi
contents Recent Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) increasingly use Reinforcement Learning (RL) for post-pretraining, such as RL with Verifiable Rewards (RLVR) for objective tasks and RL from Human Feedback (RLHF) for subjective tasks. However, RLHF incurs high costs and potential reward-policy mismatch due to reliance on human preferences, while RLVR still wastes supervision by discarding rollouts and correctness signals after each update. To address these challenges, we introduce the Synergistic Policy And Reward Co-Evolving Framework (SPARK), an efficient, on-policy, and stable method that builds on RLVR. Instead of discarding rollouts and correctness data, SPARK recycles this valuable information to simultaneously train the model itself as a generative reward model. This auxiliary training uses a mix of objectives, such as pointwise reward score, pairwise comparison, and evaluation conditioned on further-reflection responses, to teach the model to evaluate and improve its own responses. Our process eliminates the need for a separate reward model and costly human preference data. SPARK creates a positive co-evolving feedback loop: improved reward accuracy yields better policy gradients, which in turn produce higher-quality rollouts that further refine the reward model. Our unified framework supports test-time scaling via self-reflection without external reward models and their associated costs. We show that SPARK achieves significant performance gains on multiple LLM and LVLM models and multiple reasoning, reward models, and general benchmarks. For example, SPARK-VL-7B achieves an average 9.7% gain on 7 reasoning benchmarks, 12.1% on 2 reward benchmarks, and 1.5% on 8 general benchmarks over the baselines, demonstrating robustness and broad generalization.
format Preprint
id arxiv_https___arxiv_org_abs_2509_22624
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle SPARK: Synergistic Policy And Reward Co-Evolving Framework
Liu, Ziyu
Zang, Yuhang
Ding, Shengyuan
Cao, Yuhang
Dong, Xiaoyi
Duan, Haodong
Lin, Dahua
Wang, Jiaqi
Computer Vision and Pattern Recognition
Machine Learning
Recent Large Language Models (LLMs) and Large Vision-Language Models (LVLMs) increasingly use Reinforcement Learning (RL) for post-pretraining, such as RL with Verifiable Rewards (RLVR) for objective tasks and RL from Human Feedback (RLHF) for subjective tasks. However, RLHF incurs high costs and potential reward-policy mismatch due to reliance on human preferences, while RLVR still wastes supervision by discarding rollouts and correctness signals after each update. To address these challenges, we introduce the Synergistic Policy And Reward Co-Evolving Framework (SPARK), an efficient, on-policy, and stable method that builds on RLVR. Instead of discarding rollouts and correctness data, SPARK recycles this valuable information to simultaneously train the model itself as a generative reward model. This auxiliary training uses a mix of objectives, such as pointwise reward score, pairwise comparison, and evaluation conditioned on further-reflection responses, to teach the model to evaluate and improve its own responses. Our process eliminates the need for a separate reward model and costly human preference data. SPARK creates a positive co-evolving feedback loop: improved reward accuracy yields better policy gradients, which in turn produce higher-quality rollouts that further refine the reward model. Our unified framework supports test-time scaling via self-reflection without external reward models and their associated costs. We show that SPARK achieves significant performance gains on multiple LLM and LVLM models and multiple reasoning, reward models, and general benchmarks. For example, SPARK-VL-7B achieves an average 9.7% gain on 7 reasoning benchmarks, 12.1% on 2 reward benchmarks, and 1.5% on 8 general benchmarks over the baselines, demonstrating robustness and broad generalization.
title SPARK: Synergistic Policy And Reward Co-Evolving Framework
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2509.22624