Partial Policy Gradients for RL in LLMs

• URL: http://arxiv.org/abs/2603.06138v1
• Date: Fri, 06 Mar 2026 10:47:41 GMT
• Title: Partial Policy Gradients for RL in LLMs
• Abstract summary: We propose a natural approach for modeling policy structure in policy gradients.<n>Smaller subsets represent simpler policies, which can be learned more reliably because their empirical gradient estimates are more accurate.<n>We evaluate the policies empirically on multiple persona-alignment conversational problems.
• Score: 43.33467583837307
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Reinforcement learning is a framework for learning to act sequentially in an unknown environment. We propose a natural approach for modeling policy structure in policy gradients. The key idea is to optimize for a subset of future rewards: smaller subsets represent simpler policies, which can be learned more reliably because their empirical gradient estimates are more accurate. Our approach allows for modeling and comparison of different policy classes, including full planning, greedy, K-step lookahead, and segment policies. We evaluate the policies empirically on multiple persona-alignment conversational problems. Different policies excel in different problems, reflecting their different characteristics and highlighting the importance of our studied policy class.

Related papers

• Learning Policy Representations for Steerable Behavior Synthesis [80.4542176039074]
  Given a Markov decision process (MDP), we seek to learn representations for a range of policies to facilitate behavior steering at test time.<n>We show that these representations can be approximated uniformly for a range of policies using a set-based architecture.<n>We use variational generative approach to induce a smooth latent space, and further shape it with contrastive learning so that latent distances align with differences in value functions.
  arXiv  Detail & Related papers  (2026-01-29T21:52:06Z)
• Multilinear Tensor Low-Rank Approximation for Policy-Gradient Methods in Reinforcement Learning [27.868175900131313]
  Reinforcement learning (RL) aims to estimate the action to take given a (time-varying) state.<n>This paper postulates multi-linear mappings to efficiently estimate the parameters of the RL policy.<n>We leverage the PARAFAC decomposition to design tensor low-rank policies.
  arXiv  Detail & Related papers  (2025-01-08T23:22:08Z)
• Contrastive Policy Gradient: Aligning LLMs on sequence-level scores in a supervised-friendly fashion [43.77763433288893]
  We introduce Contrastive Policy Gradient, or CoPG, a simple and mathematically principled new RL algorithm that can estimate the optimal policy even from off-policy data.<n>We show this approach to generalize the direct alignment method IPO (identity preference optimization) and classic policy gradient.<n>We experiment with the proposed CoPG on a toy bandit problem to illustrate its properties, as well as for finetuning LLMs on a summarization task.
  arXiv  Detail & Related papers  (2024-06-27T14:03:49Z)
• Learning Optimal Deterministic Policies with Stochastic Policy Gradients [62.81324245896716]
  Policy gradient (PG) methods are successful approaches to deal with continuous reinforcement learning (RL) problems. In common practice, convergence (hyper)policies are learned only to deploy their deterministic version. We show how to tune the exploration level used for learning to optimize the trade-off between the sample complexity and the performance of the deployed deterministic policy.
  arXiv  Detail & Related papers  (2024-05-03T16:45:15Z)
• Policy learning "without" overlap: Pessimism and generalized empirical Bernstein's inequality [94.89246810243053]
  This paper studies offline policy learning, which aims at utilizing observations collected a priori to learn an optimal individualized decision rule.<n>Existing policy learning methods rely on a uniform overlap assumption, i.e., the propensities of exploring all actions for all individual characteristics must be lower bounded.<n>We propose Pessimistic Policy Learning (PPL), a new algorithm that optimize lower confidence bounds (LCBs) instead of point estimates.
  arXiv  Detail & Related papers  (2022-12-19T22:43:08Z)
• Towards A Unified Policy Abstraction Theory and Representation Learning Approach in Markov Decision Processes [39.94472154078338]
  We propose a unified policy abstraction theory, containing three types of policy abstraction associated to policy features at different levels. We then generalize them to three policy metrics that quantify the distance (i.e., similarity) of policies. For the empirical study, we investigate the efficacy of the proposed policy metrics and representations, in characterizing policy difference and conveying policy generalization respectively.
  arXiv  Detail & Related papers  (2022-09-16T03:41:50Z)
• Constructing a Good Behavior Basis for Transfer using Generalized Policy Updates [63.58053355357644]
  We study the problem of learning a good set of policies, so that when combined together, they can solve a wide variety of unseen reinforcement learning tasks. We show theoretically that having access to a specific set of diverse policies, which we call a set of independent policies, can allow for instantaneously achieving high-level performance.
  arXiv  Detail & Related papers  (2021-12-30T12:20:46Z)
• Policy Evaluation Networks [50.53250641051648]
  We introduce a scalable, differentiable fingerprinting mechanism that retains essential policy information in a concise embedding. Our empirical results demonstrate that combining these three elements can produce policies that outperform those that generated the training data.
  arXiv  Detail & Related papers  (2020-02-26T23:00:27Z)

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.