Hista and Numca: Estimate State Value Effectively for LLM Reinforcement Learning

• URL: http://arxiv.org/abs/2605.29782v1
• Date: Thu, 28 May 2026 11:31:13 GMT
• Title: Hista and Numca: Estimate State Value Effectively for LLM Reinforcement Learning
• Abstract summary: Reinforcement learning (RL) refines large language models (LLMs) by directly optimizing model behavior through reward signals.<n>We introduce the State Value Estimation Benchmark (SVEB) to assess state estimation within existing RL frameworks.<n>We propose two techniques: Numca, which leverages numerical spans as gradable milestones for state value estimation, and Hista, a framework that uses LLM's hidden states as representation to weighted average disjoint rollouts and their return.
• Score: 18.614878630206487
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning (RL) refines large language models (LLMs) by directly optimizing model behavior through reward signals. While accurate state value estimation is critical for stable training in classical RL, it remains an underexplored challenge in LLM post-training. In this work, we introduce the State Value Estimation Benchmark (SVEB) to assess state estimation within existing RL frameworks and show that critics in standard approaches like PPO collapse to a coarse group-average baseline. To address this, we propose two techniques: Numca, which leverages numerical spans as gradable milestones for state value estimation, and Hista, a framework that uses LLM's hidden states as representation to weighted average disjoint rollouts and their return. Extensive experiments demonstrate that both methods yield more accurate state value estimates and enhance training performance across different RL algorithms and model sizes without incurring significant computational overhead.

Related papers

• Modularized Reinforcement Learning on LLMs: From MDP Creation to Exploration and Learning [65.72189367176186]
  This survey organizes that examination around three stages of algorithm construction.<n>We begin with MDP creation: how the reward function, state space, action space, termination condition, and discount factor are, or could be, defined for LLM training.<n>We then turn to exploration, covering temperature sampling, entropy regularization, intrinsic motivation, tree search, and curriculum learning.
  arXiv  Detail & Related papers  (2026-06-20T08:20:41Z)
• Reinforcement Learning from Denoising Feedback [22.358386875891053]
  Policy loss estimation remains a fundamental and long-standing challenge in reinforcement learning (RL) for diffusion language models (dLLMs)<n>We introduce Reinforcement Learning from Denoising Feedback (RLDF), a novel training paradigm that leverages feedback obtained from rollout and training processes to facilitate accurate and efficient policy loss estimation.<n>RLDF achieves consistent and substantial improvements in both performance and generalizability across two representative dLLM architectures, LLaDA and Dream, on multiple reasoning benchmarks.
  arXiv  Detail & Related papers  (2026-05-25T09:39:13Z)
• Off-Policy Value-Based Reinforcement Learning for Large Language Models [25.962820072445222]
  ReVal is a Bellman-update-based method that combines stepwise signals capturing internal consistency with trajectory-level signals derived from outcome verification.<n>On DeepSeek-R1-Distill-1.5B, ReVal improves training efficiency and achieves improvement of 2.7% in AIME24 and 4.5% in out-of-domain benchmark GPQA over GRPO.
  arXiv  Detail & Related papers  (2026-03-24T15:55:02Z)
• RADAR: Revealing Asymmetric Development of Abilities in MLLM Pre-training [59.493415006017635]
  Pre-trained Multi-modal Large Language Models (MLLMs) provide a knowledge-rich foundation for post-training.<n>Current evaluation relies on testing after supervised fine-tuning, which introduces laborious additional training and autoregressive decoding costs.<n>We propose RADAR, an efficient ability-centric evaluation framework for Revealing Asymmetric Development of Abilities in MLLM pRe-training.
  arXiv  Detail & Related papers  (2026-02-13T12:56:31Z)
• DFPO: Scaling Value Modeling via Distributional Flow towards Robust and Generalizable LLM Post-Training [94.568675548967]
  Training reinforcement learning (RL) systems in real-world environments remains challenging due to noisy supervision and poor out-of-domain generalization.<n>Recent distributional RL methods improve robustness by modeling values with multiple quantile points, but they still learn each quantile independently as a scalar.<n>We propose DFPO, a robust distributional RL framework that models values as continuous flows across time steps.
  arXiv  Detail & Related papers  (2026-02-05T17:07:42Z)
• A Comedy of Estimators: On KL Regularization in RL Training of LLMs [81.7906270099878]
  reinforcement learning (RL) can substantially improve the reasoning performance of large language models (LLMs)<n>The RL objective for LLM training involves a regularization term, which is the reverse Kullback-Leibler (KL) divergence between the trained policy and the reference policy.<n>Recent works show that prevailing practices for incorporating KL regularization do not provide correct gradients for stated objectives, creating a discrepancy between the objective and its implementation.<n>We study the gradients of several estimators configurations, revealing how design choices shape gradient bias.
  arXiv  Detail & Related papers  (2025-12-26T04:20:58Z)
• Lean and Mean: Decoupled Value Policy Optimization with Global Value Guidance [52.65461207786633]
  Policy-based Reinforcement Learning from Human Feedback is essential for aligning large language models with human preferences.<n>It requires joint training of an actor and critic with a pretrained, fixed reward model for guidance.<n>We propose textbfDecoupled Value Policy Optimization (DVPO), a lean framework that replaces traditional reward modeling with a pretrained emphglobal value model (GVM)
  arXiv  Detail & Related papers  (2025-02-24T08:11:33Z)
• VinePPO: Refining Credit Assignment in RL Training of LLMs [66.80143024475635]
  We propose VinePPO, a straightforward approach that leverages the flexibility of language environments to compute unbiased Monte Carlo-based estimates.<n>Our method consistently outperforms PPO and other baselines across MATH and GSM8K datasets in less wall-clock time.
  arXiv  Detail & Related papers  (2024-10-02T15:49:30Z)
• Value-Consistent Representation Learning for Data-Efficient Reinforcement Learning [105.70602423944148]
  We propose a novel method, called value-consistent representation learning (VCR), to learn representations that are directly related to decision-making. Instead of aligning this imagined state with a real state returned by the environment, VCR applies a $Q$-value head on both states and obtains two distributions of action values. It has been demonstrated that our methods achieve new state-of-the-art performance for search-free RL algorithms.
  arXiv  Detail & Related papers  (2022-06-25T03:02:25Z)
• Value Gradient weighted Model-Based Reinforcement Learning [28.366157882991565]
  Model-based reinforcement learning (MBRL) is a sample efficient technique to obtain control policies. VaGraM is a novel method for value-aware model learning.
  arXiv  Detail & Related papers  (2022-04-04T13:28:31Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.