Stabilizing Efficient Reasoning with Step-Level Advantage Selection

• URL: http://arxiv.org/abs/2604.24003v1
• Date: Mon, 27 Apr 2026 03:34:59 GMT
• Title: Stabilizing Efficient Reasoning with Step-Level Advantage Selection
• Abstract summary: We show that short-context post-training alone, using standard GRPO without any length-aware objective, already induces substantial reasoning compression.<n>We propose Step-level Advantage Selection (SAS), which operates at the reasoning-step level and assigns a zero advantage to low-confidence steps in correct rollouts and to high-confidence steps in verifier-failed rollouts.
• Score: 54.563811052329235
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models (LLMs) achieve strong reasoning performance by allocating substantial computation at inference time, often generating long and verbose reasoning traces. While recent work on efficient reasoning reduces this overhead through length-based rewards or pruning, many approaches are post-trained under a much shorter context window than base-model training, a factor whose effect has not been systematically isolated. We first show that short-context post-training alone, using standard GRPO without any length-aware objective, already induces substantial reasoning compression-but at the cost of increasingly unstable training dynamics and accuracy degradation. To address this, we propose Step-level Advantage Selection (SAS), which operates at the reasoning-step level and assigns a zero advantage to low-confidence steps in correct rollouts and to high-confidence steps in verifier-failed rollouts, where failures often arise from truncation or verifier issues rather than incorrect reasoning. Across diverse mathematical and general reasoning benchmarks, SAS improves average Pass@1 accuracy by 0.86 points over the strongest length-aware baseline while reducing average reasoning length by 16.3%, yielding a better accuracy-efficiency trade-off.

Related papers

• SABER: Stability-Aware Early Exit for LLM Reasoning via Adversarial Branch Probing [32.867599324353456]
  SABER is a training-free framework for stability-aware early exit via adversarial branch probing.<n>It reduces reasoning token consumption by 30.2%--39.8% on average while maintaining competitive accuracy with full-length reasoning.
  arXiv  Detail & Related papers  (2026-08-28T06:13:23Z)
• ChainPrune: Evaluating and Reducing Redundancy in Long Chain-of-Thought Reasoning [48.3829293800053]
  We propose ChainPrune, a novel reasoning path semantic structural optimization method.<n>We consolidate self-generated reasoning paths into a tree-based structure, followed by a multi-criteria dominant path selection process.<n>We incorporate a DPO-based preference learning method combined with supervised loss, effectively mitigating false reward suppression.
  arXiv  Detail & Related papers  (2026-08-22T09:09:05Z)
• CARE: Competence-Aware Reward Shaping for Adaptive Reasoning Length in Video-MLLMs [50.189987475377656]
  We propose CARE, a competence-aware reward shaping framework for adaptive reasoning length optimization in multimodal reasoning.<n> CARE maintains a smoothed competence estimate via an exponential moving average of pass rates, and uses it to route training into progressive stages.<n>Experiments on multiple video reasoning and general video understanding benchmarks demonstrate that CARE consistently improves reasoning accuracy, stabilizes reinforcement learning, and significantly enhances token efficiency.
  arXiv  Detail & Related papers  (2026-06-18T08:28:26Z)
• Correct Answers from Sound Reasoning: Verifiable Process Supervision for Language Models [94.68358825189738]
  We propose a post-training framework for verifiable domains that jointly optimize prediction accuracy and reasoning quality.<n>We evaluate VPS on chess, a controlled testbed where reasoning steps can be deterministically verified against engine signals.<n>VPS preserves accuracy while significantly improving reasoning quality, reducing win-rate error by up to 30% and restoring consistency to near saturation.
  arXiv  Detail & Related papers  (2026-04-03T15:19:46Z)
• Learning When to Sample: Confidence-Aware Self-Consistency for Efficient LLM Chain-of-Thought Reasoning [7.8668388431725695]
  Large language models (LLMs) achieve strong reasoning performance through chain-of-thought (CoT) reasoning.<n>Recent self-consistency-based approaches further improve accuracy but require sampling and aggregating multiple reasoning trajectories.<n>This paper introduces a confidence-aware decision framework that analyzes a single completed reasoning trajectory to adaptively select between single-path and multi-path reasoning.
  arXiv  Detail & Related papers  (2026-03-09T22:34:06Z)
• Stepwise Penalization for Length-Efficient Chain-of-Thought Reasoning [66.22060690012512]
  Large reasoning models improve with more test-time computation, but often overthink, producing unnecessarily long chains-of-thought that raise cost without improving accuracy.<n>We propose Step-wise Adaptive Penalization (SWAP), a fine-grained framework that allocates length reduction across steps based on intrinsic contribution.
  arXiv  Detail & Related papers  (2026-02-27T20:23:59Z)
• WS-GRPO: Weakly-Supervised Group-Relative Policy Optimization for Rollout-Efficient Reasoning [67.45237332694025]
  Group Relative Policy Optimization is effective for training language models on complex reasoning.<n>We propose Weakly Supervised GRPO, which improves rollout efficiency by converting terminal rewards into correctness-aware guidance.
  arXiv  Detail & Related papers  (2026-02-19T02:43:35Z)
• Constraint-Rectified Training for Efficient Chain-of-Thought [60.52883907721588]
  Chain-of-Thought (CoT) has significantly enhanced the reasoning capabilities of Large Language Models (LLMs)<n>While longer reasoning traces can improve answer quality and unlock abilities such as self-correction, they also incur high inference costs and often introduce redundant steps, known as overthinking.<n>Recent research seeks to develop efficient reasoning strategies that balance reasoning length and accuracy.
  arXiv  Detail & Related papers  (2026-02-13T02:13:45Z)
• 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)
• Think or Not? Exploring Thinking Efficiency in Large Reasoning Models via an Information-Theoretic Lens [51.90059610606049]
  This paper revisits the efficiency of such reasoning processes through an information-theoretic lens.<n>We propose two metrics, InfoBias and InfoGain, to quantify divergence from ideal reasoning paths and stepwise information contribution.<n>Motivated by these findings, we introduce an entropy-based Adaptive Think strategy that dynamically halts reasoning once confidence is sufficiently high.
  arXiv  Detail & Related papers  (2025-05-23T13:38:56Z)
• ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning [64.93140713419561]
  Large Reasoning Models (LRMs) perform strongly in complex reasoning tasks via Chain-of-Thought (CoT) prompting, but often suffer from verbose outputs.<n>Existing fine-tuning-based compression methods either operate post-hoc pruning, risking disruption to reasoning coherence, or rely on sampling-based selection.<n>We introduce ConCISE, a framework designed to generate concise reasoning chains, integrating Confidence Injection to boost reasoning confidence, and Early Stopping to terminate reasoning when confidence is sufficient.
  arXiv  Detail & Related papers  (2025-05-08T01:40:40Z)

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.