Stable Reinforcement Learning for Efficient Reasoning

• URL: http://arxiv.org/abs/2505.18086v1
• Date: Fri, 23 May 2025 16:43:03 GMT
• Title: Stable Reinforcement Learning for Efficient Reasoning
• Authors: Muzhi Dai, Shixuan Liu, Qingyi Si,
• Abstract summary: GRPO-$lambda$ is an efficient and stabilized variant of GRPO.<n>It dynamically adjusts the reward strategy by monitoring the correctness ratio.<n>It improves average accuracy by 1.48% while reducing CoT sequence length by 47.3%.
• Score: 2.838966689544288
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: The success of Deepseek-R1 has drawn the LLM community's attention to reinforcement learning (RL) methods like GRPO. However, such rule-based 0/1 outcome reward methods lack the capability to regulate the intermediate reasoning processes during chain-of-thought (CoT) generation, leading to severe overthinking phenomena. In response, recent studies have designed reward functions to reinforce models' behaviors in producing shorter yet correct completions. Nevertheless, we observe that these length-penalty reward functions exacerbate RL training instability: as the completion length decreases, model accuracy abruptly collapses, often occurring early in training. To address this issue, we propose a simple yet effective solution GRPO-$\lambda$, an efficient and stabilized variant of GRPO, which dynamically adjusts the reward strategy by monitoring the correctness ratio among completions within each query-sampled group. A low correctness ratio indicates the need to avoid length penalty that compromises CoT quality, triggering a switch to length-agnostic 0/1 rewards that prioritize reasoning capability. A high ratio maintains length penalties to boost efficiency. Experimental results show that our approach avoids training instability caused by length penalty while maintaining the optimal accuracy-efficiency trade-off. On the GSM8K, GPQA, MATH-500, AMC 2023, and AIME 2024 benchmarks, it improves average accuracy by 1.48% while reducing CoT sequence length by 47.3%.

Related papers

• Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty [59.97939500426759]
  This paper describes RLCR, an approach to training reasoning models that jointly improves accuracy and confidence estimation.<n>We show that across diverse datasets, RLCR substantially improves calibration with no loss in accuracy.<n>We also demonstrate that verbalized confidence can be leveraged at test time to improve accuracy and calibration.
  arXiv  Detail & Related papers  (2025-07-22T17:56:01Z)
• Scaling Up RL: Unlocking Diverse Reasoning in LLMs via Prolonged Training [121.5858973157225]
  We investigate the effects of prolonged reinforcement learning on a small language model across a diverse set of reasoning domains.<n>We introduce controlled KL regularization, clipping ratio, and periodic reference policy resets as critical components for unlocking long-term performance gains.<n>Our model achieves significant improvements over strong baselines, including +14.7% on math, +13.9% on coding, and +54.8% on logic puzzle tasks.
  arXiv  Detail & Related papers  (2025-07-16T17:59:24Z)
• AALC: Large Language Model Efficient Reasoning via Adaptive Accuracy-Length Control [18.273777938294327]
  Large reasoning models (LRMs) achieve impressive reasoning capabilities by generating lengthy chain-of-thoughts.<n>We introduce AALC, a lightweight, accuracy-aware length reward integrated into reinforcement learning.<n>We show that our approach reduces response length by over 50% while maintaining or even improving the original accuracy.
  arXiv  Detail & Related papers  (2025-06-25T06:29:18Z)
• GRPO-CARE: Consistency-Aware Reinforcement Learning for Multimodal Reasoning [53.894789613838654]
  We introduce SEED-Bench-R1, a benchmark with complex real-world videos requiring balanced perception and reasoning.<n>Using SEED-Bench-R1, we find that standard GRPO, while improving answer accuracy, often reduces logical coherence between reasoning steps and answers, with only a 57.9% consistency rate.<n>We propose GRPO-CARE, a consistency-aware RL framework optimizing both answer correctness and reasoning coherence without explicit supervision.
  arXiv  Detail & Related papers  (2025-06-19T08:49:13Z)
• Thinking Fast and Right: Balancing Accuracy and Reasoning Length with Adaptive Rewards [17.829990749622496]
  We propose an adaptive reward-shaping method for large language models.<n>Our method dynamically adjusts the trade-off between accuracy and response length based on model performance.<n> Experiments show that our approach consistently and dramatically reduces reasoning length while largely maintaining accuracy.
  arXiv  Detail & Related papers  (2025-05-23T18:44:46Z)
• Learn to Reason Efficiently with Adaptive Length-based Reward Shaping [23.626013831589212]
  Large Reasoning Models (LRMs) have shown remarkable capabilities in solving complex problems through reinforcement learning (RL)<n>We present a unified framework that formulates various efficient reasoning methods through the lens of length-based reward shaping.<n>Experiments on DeepSeek-R1-Distill-Qwen-1.5B, DeepSeek-R1-Distill-Qwen-7B, and DeepSeek-R1-Distill-Qwen-32B show that our approach significantly enhances both reasoning performance and response length efficiency.
  arXiv  Detail & Related papers  (2025-05-21T15:03:26Z)
• S-GRPO: Early Exit via Reinforcement Learning in Reasoning Models [2.9925837108958864]
  Test-Time Scaling emerges as an active research focus in the large language model community.<n>Recent studies reveal that reasoning models (even Qwen3) consistently exhibit excessive thought redundancy.<n>This paper introduces Serial-Group Decaying-Reward Policy Optimization (S-GRPO), a novel reinforcement learning paradigm.
  arXiv  Detail & Related papers  (2025-05-12T15:50:44Z)
• Dynamic Early Exit in Reasoning Models [13.982812528756504]
  Overthinking in long chain-of-thought (CoT) generation slows down the efficiency of problem solving, but also risks accuracy loss.<n>We propose a simple yet effective method that allows LLMs to self-truncate CoT sequences by early exit during generation.<n>Our method requires no additional training and can be seamlessly integrated into existing o1-like reasoning LLMs.
  arXiv  Detail & Related papers  (2025-04-22T13:36:53Z)
• A Minimalist Approach to LLM Reasoning: from Rejection Sampling to Reinforce [68.99924691391048]
  We revisit GRPO from a reinforce-like algorithm perspective and analyze its core components.<n>We find that a simple rejection sampling baseline, RAFT, yields competitive performance than GRPO and PPO.<n>Motivated by this insight, we propose Reinforce-Rej, a minimal extension of policy gradient that filters both entirely incorrect and entirely correct samples.
  arXiv  Detail & Related papers  (2025-04-15T16:15:02Z)
• Exploring the Limit of Outcome Reward for Learning Mathematical Reasoning [65.2421542320293]
  Reasoning abilities are crucial components of general intelligence.<n>Recent advances by proprietary companies, such as o-series models of OpenAI, have made remarkable progress on reasoning tasks.<n>This paper proposes a new RL framework, termed OREAL, to pursue the performance limit that can be achieved through textbfOutcome textbfREwtextbfArd-based reinforcement textbfLearning for mathematical reasoning tasks.
  arXiv  Detail & Related papers  (2025-02-10T18:57:29Z)
• Self-Consistency Preference Optimization [79.37880123635405]
  We introduce self-consistency preference optimization (ScPO) ScPO iteratively trains consistent answers to be preferred over inconsistent ones on unsupervised new problems. On ZebraLogic, ScPO fine Llamatunes-3 8B to be superior to Llama-3 70B, Gemma-2 27B, and Claude-3 Haiku.
  arXiv  Detail & Related papers  (2024-11-06T18:36:22Z)
• LoRanPAC: Low-rank Random Features and Pre-trained Models for Bridging Theory and Practice in Continual Learning [103.45785408116146]
  Continual learning (CL) aims to train a model that can solve multiple tasks presented sequentially.<n>Recent CL approaches have achieved strong performance by leveraging large pre-trained models that generalize well to downstream tasks.<n>However, such methods lack theoretical guarantees, making them prone to unexpected failures.<n>We aim to bridge this gap by designing a simple CL method that is theoretically sound and highly performant.
  arXiv  Detail & Related papers  (2024-10-01T12:58:37Z)
• WARM: On the Benefits of Weight Averaged Reward Models [63.08179139233774]
  We propose Weight Averaged Reward Models (WARM) to mitigate reward hacking. Experiments on summarization tasks, using best-of-N and RL methods, shows that WARM improves the overall quality and alignment of LLM predictions.
  arXiv  Detail & Related papers  (2024-01-22T18:27:08Z)
• Low-Precision Reinforcement Learning [63.930246183244705]
  Low-precision training has become a popular approach to reduce computation time, memory footprint, and energy consumption in supervised learning. In this paper we consider continuous control with the state-of-the-art SAC agent and demonstrate that a na"ive adaptation of low-precision methods from supervised learning fails.
  arXiv  Detail & Related papers  (2021-02-26T16:16:28Z)

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.