Future Policy Aware Preference Learning for Mathematical Reasoning

• URL: http://arxiv.org/abs/2509.19893v1
• Date: Wed, 24 Sep 2025 08:44:12 GMT
• Title: Future Policy Aware Preference Learning for Mathematical Reasoning
• Abstract summary: A key challenge is the large token overlap between preferred and dispreferred trajectories.<n>We propose Future Policy Aware (FPA) preference learning, which replaces the current policy with a future policy in the regularization term.<n>FPA yields consistent performance gains, with the largest improvements observed with SimPER, achieving gains of up to 5.75%.
• Score: 15.444539171776983
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Preference learning methods such as Direct Preference Optimization (DPO) have become standard for Large Language Model (LLM) post-training, yet they are often ineffective for mathematical reasoning. A key challenge is the large token overlap between preferred and dispreferred trajectories; lowering the probability of dispreferred trajectories also reduces the probability of shared useful tokens, leading to over-penalization and overall performance collapse. As a mitigation, existing algorithms include the probability of a trajectory under the current policy as a regularization term, which decreases the effect of the gradient when the probability is low. However, by the time this effect takes hold, useful tokens may have already been over-penalized as the model has begun to degrade. To address this, we propose Future Policy Aware (FPA) preference learning, which replaces the current policy with a future policy in the regularization term. This future policy is estimated via lightweight, logit-space extrapolation from a reference model toward the current model. FPA enables safer training by preemptively regularizing potentially problematic gradients. We apply FPA to DPO, RPO, and SimPER and evaluate them on the MATH and GSM8K benchmarks. FPA yields consistent performance gains, with the largest improvements observed with SimPER, achieving gains of up to 5.75%. We demonstrate that FPA provides proactive regularization while preserving the probability of shared, useful mathematical tokens, and enables longer, degradation-free training with negligible computational overhead. We will release our code publicly upon publication.

Related papers

• VIMPO: Value-Implicit Policy Optimization for LLMs [106.88933849641272]
  Group-relative methods such as GRPO avoid training a critic, but typically assign a trajectory-level advantage to every token.<n>Actor-critic methods provide denser learning signals, but require a learned value function with its own training instability.<n>We introduce VIMPO, a critic-free policy optimization method that derives a policy-implied value function from optimality conditions of KL-regularized reinforcement learning.
  arXiv  Detail & Related papers  (2026-06-18T09:44:12Z)
• One-Way Policy Optimization for Self-Evolving LLMs [63.8638342097375]
  Reinforcement Learning with Verifiable Rewards (RLVR) has become a promising paradigm for scaling reasoning capabilities of Large Language Models (LLMs)<n>We propose One-Way Policy Optimization (OWPO), a method based on the principle of decoupling optimization direction from update magnitude.<n> Experimental results demonstrate that OWPO outperforms strong baselines, including DAPO, OPD, and MOPD.
  arXiv  Detail & Related papers  (2026-05-21T08:25:27Z)
• Stochastic MeanFlow Policies: One-Step Generative Control with Entropic Mirror Descent [53.828537014796574]
  Online off-policy reinforcement learning (RL) is shaped by two choices: the policy class and the update rule.<n>We propose MeanFlow Policies, a one-step generative policy class that maps noise to actions through a MeanFlow transformation.<n>Across seven MuJoCo benchmarks, S improves over Gaussian and generative baselines while retaining single-step inference efficiency.
  arXiv  Detail & Related papers  (2026-05-20T15:14:14Z)
• Beyond Negative Rollouts: Positive-Only Policy Optimization with Implicit Negative Gradients [19.444308344942716]
  Reinforcement learning with verifiable rewards (RLVR) becomes a dominant paradigm for enhancing the reasoning ability of large language models (LLMs)<n>We propose Positive-Only Policy Optimization (POPO), a novel RLVR framework in which learning can occur exclusively via online positive rollouts.
  arXiv  Detail & Related papers  (2026-05-07T17:55:21Z)
• Rethinking the Trust Region in LLM Reinforcement Learning [72.25890308541334]
  Proximal Policy Optimization (PPO) serves as the de facto standard algorithm for Large Language Models (LLMs)<n>We propose Divergence Proximal Policy Optimization (DPPO), which substitutes clipping with a more principled constraint.<n>DPPO achieves superior training and efficiency compared to existing methods, offering a more robust foundation for RL-based fine-tuning.
  arXiv  Detail & Related papers  (2026-02-04T18:59:04Z)
• Ratio-Variance Regularized Policy Optimization for Efficient LLM Fine-tuning [48.34492357368989]
  We propose a primal-dual framework that supports stable on-policy learning and enables principled off-policy data reuse.<n>$R2VPO$ achieves superior performance with average relative gains of up to 17% over strong clipping-based baselines.
  arXiv  Detail & Related papers  (2026-01-06T14:01:42Z)
• Mind the Gap: Data Rewriting for Stable Off-Policy Supervised Fine-Tuning [33.899779762210976]
  Supervised fine-tuning (SFT) of large language models can be viewed as an off-policy learning problem.<n>Existing methods mitigate this issue with KL penalties or clipping, which passively updates rather than actively reducing the gap.<n>We propose a simple yet effective data rewriting framework that proactively shrinks the policy gap before training.
  arXiv  Detail & Related papers  (2025-09-18T17:02:30Z)
• GTPO: Trajectory-Based Policy Optimization in Large Language Models [42.60363805227946]
  Policy-based optimizations are widely adopted today for the training and alignment of language models.<n>In this paper, we reveal and analyze two major limitations of GRPO.<n>We introduce GTPO, which identifies conflict tokens, tokens appearing in the same position across completions with opposite rewards.
  arXiv  Detail & Related papers  (2025-08-05T08:15:01Z)
• Geometric-Mean Policy Optimization [117.05113769757172]
  Group Relative Policy Optimization ( GRPO) has significantly enhanced the reasoning capability of large language models.<n> GRPO is observed to suffer from unstable policy updates when facing tokens with outlier importance-weighted rewards.<n>We propose Geometric-Mean Policy Optimization (GMPO) to improve the stability of GRPO through suppressing token reward outliers.
  arXiv  Detail & Related papers  (2025-07-28T09:54:05Z)
• Reusing Trajectories in Policy Gradients Enables Fast Convergence [59.27926064817273]
  Policy gradient (PG) methods are a class of effective reinforcement learning algorithms.<n>We propose RPG (Retrospective Policy Gradient), a PG algorithm that combines old and new trajectories for policy updates.<n>Under established assumptions, RPG achieves a sample complexity of $widetildeO(epsilon-1)$, the best known rate in the literature.
  arXiv  Detail & Related papers  (2025-06-06T15:42:15Z)
• Enhancing PPO with Trajectory-Aware Hybrid Policies [6.938941097426891]
  Proximal policy optimization (PPO) is one of the most popular state-of-the-art on-policy algorithms.<n>High variance, and high sample complexity still remain critical challenges in on-policy algorithms.<n>We propose Hybrid-Policy Proximal Policy Optimization (HP3O), which utilizes a trajectory replay buffer to make efficient use of trajectories generated by recent policies.
  arXiv  Detail & Related papers  (2025-02-21T22:00:13Z)
• Policy Gradient with Active Importance Sampling [55.112959067035916]
  Policy gradient (PG) methods significantly benefit from IS, enabling the effective reuse of previously collected samples. However, IS is employed in RL as a passive tool for re-weighting historical samples. We look for the best behavioral policy from which to collect samples to reduce the policy gradient variance.
  arXiv  Detail & Related papers  (2024-05-09T09:08:09Z)
• Off-Policy Evaluation for Large Action Spaces via Policy Convolution [60.6953713877886]
  Policy Convolution family of estimators uses latent structure within actions to strategically convolve the logging and target policies. Experiments on synthetic and benchmark datasets demonstrate remarkable mean squared error (MSE) improvements when using PC.
  arXiv  Detail & Related papers  (2023-10-24T01:00:01Z)
• Provably Good Batch Reinforcement Learning Without Great Exploration [51.51462608429621]
  Batch reinforcement learning (RL) is important to apply RL algorithms to many high stakes tasks. Recent algorithms have shown promise but can still be overly optimistic in their expected outcomes. We show that a small modification to Bellman optimality and evaluation back-up to take a more conservative update can have much stronger guarantees.
  arXiv  Detail & Related papers  (2020-07-16T09:25:54Z)
• Preference-based Reinforcement Learning with Finite-Time Guarantees [76.88632321436472]
  Preference-based Reinforcement Learning (PbRL) replaces reward values in traditional reinforcement learning to better elicit human opinion on the target objective. Despite promising results in applications, the theoretical understanding of PbRL is still in its infancy. We present the first finite-time analysis for general PbRL problems.
  arXiv  Detail & Related papers  (2020-06-16T03:52:41Z)
• Optimizing for the Future in Non-Stationary MDPs [52.373873622008944]
  We present a policy gradient algorithm that maximizes a forecast of future performance. We show that our algorithm, called Prognosticator, is more robust to non-stationarity than two online adaptation techniques.
  arXiv  Detail & Related papers  (2020-05-17T03:41:19Z)

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.