Research

Paper

TESTING March 11, 2026

$V_{0.5}$: Generalist Value Model as a Prior for Sparse RL Rollouts

Authors

Yi-Kai Zhang, Yueqing Sun, Hongyan Hao, Qi Gu, Xunliang Cai, De-Chuan Zhan, Han-Jia Ye

Abstract

In Reinforcement Learning with Verifiable Rewards (RLVR), constructing a robust advantage baseline is critical for policy gradients, effectively guiding the policy model to reinforce desired behaviors. Recent research has introduced Generalist Value Models (such as $V_0$), which achieve pre-trained value estimation by explicitly encoding model capabilities in-context, eliminating the need to synchronously update the value model alongside the policy model. In this paper, we propose $V_{0.5}$, which adaptively fuses the baseline predicted by such value model (acting as a prior) with the empirical mean derived from sparse rollouts. This constructs a robust baseline that balances computational efficiency with extremely low variance. Specifically, we introduce a real-time statistical testing and dynamic budget allocation. This balances the high variance caused by sparse sampling against the systematic bias (or hallucinations) inherent in the value model's prior. By constructing a hypothesis test to evaluate the prior's reliability in real-time, the system dynamically allocates additional rollout budget on demand. This mechanism minimizes the baseline estimator's Mean Squared Error (MSE), guaranteeing stable policy gradients, even under extreme sparsity with a group size of 4. Extensive evaluations across six mathematical reasoning benchmarks demonstrate that $V_{0.5}$ significantly outperforms GRPO and DAPO, achieving faster convergence and over some 10% performance improvement.

Metadata

arXiv ID: 2603.10848
Provider: ARXIV
Primary Category: cs.LG
Published: 2026-03-11
Fetched: 2026-03-12 04:21

Related papers

Raw Data (Debug)
{
  "raw_xml": "<entry>\n    <id>http://arxiv.org/abs/2603.10848v1</id>\n    <title>$V_{0.5}$: Generalist Value Model as a Prior for Sparse RL Rollouts</title>\n    <updated>2026-03-11T14:57:41Z</updated>\n    <link href='https://arxiv.org/abs/2603.10848v1' rel='alternate' type='text/html'/>\n    <link href='https://arxiv.org/pdf/2603.10848v1' rel='related' title='pdf' type='application/pdf'/>\n    <summary>In Reinforcement Learning with Verifiable Rewards (RLVR), constructing a robust advantage baseline is critical for policy gradients, effectively guiding the policy model to reinforce desired behaviors. Recent research has introduced Generalist Value Models (such as $V_0$), which achieve pre-trained value estimation by explicitly encoding model capabilities in-context, eliminating the need to synchronously update the value model alongside the policy model. In this paper, we propose $V_{0.5}$, which adaptively fuses the baseline predicted by such value model (acting as a prior) with the empirical mean derived from sparse rollouts. This constructs a robust baseline that balances computational efficiency with extremely low variance. Specifically, we introduce a real-time statistical testing and dynamic budget allocation. This balances the high variance caused by sparse sampling against the systematic bias (or hallucinations) inherent in the value model's prior. By constructing a hypothesis test to evaluate the prior's reliability in real-time, the system dynamically allocates additional rollout budget on demand. This mechanism minimizes the baseline estimator's Mean Squared Error (MSE), guaranteeing stable policy gradients, even under extreme sparsity with a group size of 4. Extensive evaluations across six mathematical reasoning benchmarks demonstrate that $V_{0.5}$ significantly outperforms GRPO and DAPO, achieving faster convergence and over some 10% performance improvement.</summary>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.LG'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.AI'/>\n    <category scheme='http://arxiv.org/schemas/atom' term='cs.CL'/>\n    <published>2026-03-11T14:57:41Z</published>\n    <arxiv:primary_category term='cs.LG'/>\n    <author>\n      <name>Yi-Kai Zhang</name>\n    </author>\n    <author>\n      <name>Yueqing Sun</name>\n    </author>\n    <author>\n      <name>Hongyan Hao</name>\n    </author>\n    <author>\n      <name>Qi Gu</name>\n    </author>\n    <author>\n      <name>Xunliang Cai</name>\n    </author>\n    <author>\n      <name>De-Chuan Zhan</name>\n    </author>\n    <author>\n      <name>Han-Jia Ye</name>\n    </author>\n  </entry>"
}