Rethinking RL for LLM Reasoning: It's Sparse Policy Selection, Not Capability Learning

• URL: http://arxiv.org/abs/2605.06241v2
• Date: Fri, 08 May 2026 19:48:19 GMT
• Title: Rethinking RL for LLM Reasoning: It's Sparse Policy Selection, Not Capability Learning
• Abstract summary: We show thatReinforcement learning does not teach new strategies; it redistributes probability mass over solutions the base model already contains.<n>We translate this insight into ReasonMaxxer, a minimal RL-free method that applies contrastive loss only at entropy-gated decision points.
• Score: 16.73341451930523
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reinforcement learning has become the standard for improving reasoning in large language models, yet evidence increasingly suggests that RL does not teach new strategies; it redistributes probability mass over solutions the base model already contains. In this work, we ask: if RL merely steers the model toward paths it already knows, is the RL optimization loop itself necessary? Through token-level analysis across multiple model families and RL algorithms, we find that RL's beneficial footprint is a sparse, predictable correction concentrated at high-entropy decision points where the model is uncertain which branch to take. Only 1--3\% of token positions are affected, the promoted token always lies within the base model's top-5 alternatives, and targeted corrections at those few positions causally recover a large fraction of RL's accuracy gain, while random corrections fail. The base model's own entropy identifies these positions without any RL-trained model, and the entire correction is low-dimensional, representable in a tiny fraction of model parameters. These findings reframe reasoning improvement as sparse policy selection, not capability acquisition. We translate this insight into ReasonMaxxer, a minimal RL-free method that applies contrastive loss only at entropy-gated decision points, using a few hundred base-model rollouts and no online generation. Across three model families, six scales, and six math reasoning benchmarks, ReasonMaxxer matches or exceeds full RL performance while requiring only tens of problems and minutes of single-GPU training, a reduction in training cost of roughly three orders of magnitude.

Related papers

• Save the Good Prefix: Precise Error Penalization via Process-Supervised RL to Enhance LLM Reasoning [59.76691952347156]
  Reinforcement learning (RL) has emerged as a powerful framework for improving the reasoning capabilities of large language models (LLMs)<n>Most existing RL approaches rely on sparse outcome rewards, which fail to credit correct intermediate steps in partially successful solutions.<n>We propose Verifiable Prefix Policy Optimization (VPPO), which uses PRMs only to localize the first error during RL.
  arXiv  Detail & Related papers  (2026-01-26T21:38:20Z)
• InT: Self-Proposed Interventions Enable Credit Assignment in LLM Reasoning [32.274434679047395]
  Outcome-reward reinforcement learning (RL) has proven effective at improving the reasoning capabilities of large language models (LLMs)<n>Standard RL assigns credit only at the level of the final answer, penalizing entire reasoning traces when the outcome is incorrect.<n>We introduce Intervention Training (InT), a training paradigm in which the model performs fine-grained credit assignment on its own reasoning traces.
  arXiv  Detail & Related papers  (2026-01-20T18:15:38Z)
• Nudging the Boundaries of LLM Reasoning [77.26972440427285]
  Current online reinforcement learning algorithms cannot learn from problems that are "unsolvable" to the model.<n>We propose NuRL, a "nudging" method that aims to push the upper bound of LLM reasoning using self-generated hints.<n>NuRL achieves consistent improvements across 6 benchmarks and 3 models, while remaining complementary to test-time scaling.
  arXiv  Detail & Related papers  (2025-09-30T02:01:40Z)
• RL's Razor: Why Online Reinforcement Learning Forgets Less [13.481023080238534]
  We compare fine-tuning models with reinforcement learning (RL) and supervised fine-tuning (SFT)<n>We find that the degree of forgetting is determined by the distributional shift, measured as the KL-divergence between the fine-tuned and base policy evaluated on the new task.
  arXiv  Detail & Related papers  (2025-09-04T14:38:08Z)
• Reinforcement Learning Fine-Tunes a Sparse Subnetwork in Large Language Models [0.0]
  It is often assumed thatReinforcement learning (RL) fine-tuning requires updating most of a model's parameters.<n>We call this phenomenon RL-induced parameter update sparsity.<n>We show that fine-tuning only this sparse subnetwork recovers full model performance and yields parameters nearly identical to the fully fine-tuned model.
  arXiv  Detail & Related papers  (2025-07-23T01:02:17Z)
• Beyond Accuracy: Dissecting Mathematical Reasoning for LLMs Under Reinforcement Learning [93.00629872970364]
  Reinforcement learning (RL) has become the dominant paradigm for improving the performance of language models on complex reasoning tasks.<n>We introduce SPARKLE, a fine-grained analytic framework to dissect the effects of RL across three key dimensions.<n>We study whether difficult problems -- those yielding no RL signals and mixed-quality reasoning traces -- can still be effectively used for training.
  arXiv  Detail & Related papers  (2025-06-05T07:53:59Z)
• ProRL: Prolonged Reinforcement Learning Expands Reasoning Boundaries in Large Language Models [89.37819814048288]
  We introduce ProRL, a novel training methodology that incorporates KL divergence control, reference policy, and a diverse suite of tasks.<n>Our empirical analysis reveals that RL-trained models consistently outperform base resetting models across a wide range of pass@k evaluations.<n>These findings offer new insights into the conditions under which RL meaningfully expands reasoning boundaries in language models.
  arXiv  Detail & Related papers  (2025-05-30T17:59:01Z)
• RAST: Reasoning Activation in LLMs via Small-model Transfer [33.32587030836428]
  Reinforcement learning (RL) has become a powerful approach for improving the reasoning capabilities of large language models (LLMs)<n>Applying RL at scale remains intimidatingly resource-intensive, requiring multiple model copies and extensive GPU workloads.<n>We propose RAST, a simple yet effective method that transfers reasoning behaviors by injecting RL-induced probability adjustments from a small RL-trained model into larger models.
  arXiv  Detail & Related papers  (2025-05-30T17:57:08Z)
• Does Reinforcement Learning Really Incentivize Reasoning Capacity in LLMs Beyond the Base Model? [66.61292196146016]
  Reinforcement Learning with Verifiable Rewards (RLVR) has recently demonstrated notable success in enhancing the reasoning performance of large language models (LLMs)<n>This study critically examines the current state of RLVR.<n>We find that the current training setup does not elicit fundamentally new reasoning patterns.
  arXiv  Detail & Related papers  (2025-04-18T17:59:56Z)
• Improving Large Language Models via Fine-grained Reinforcement Learning with Minimum Editing Constraint [104.53687944498155]
  Reinforcement learning (RL) has been widely used in training large language models (LLMs) We propose a new RL method named RLMEC that incorporates a generative model as the reward model. Based on the generative reward model, we design the token-level RL objective for training and an imitation-based regularization for stabilizing RL process.
  arXiv  Detail & Related papers  (2024-01-11T17:58:41Z)
• Oracle Inequalities for Model Selection in Offline Reinforcement Learning [105.74139523696284]
  We study the problem of model selection in offline RL with value function approximation. We propose the first model selection algorithm for offline RL that achieves minimax rate-optimal inequalities up to logarithmic factors. We conclude with several numerical simulations showing it is capable of reliably selecting a good model class.
  arXiv  Detail & Related papers  (2022-11-03T17:32:34Z)

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.