Breaking the Exploration Bottleneck: Rubric-Scaffolded Reinforcement
  Learning for General LLM Reasoning

Yang Zhou; Sunzhu Li; Shunyu Liu; Wenkai Fang; Jiale Zhao; Jingwen Yang; Jianwei Lv; Kongcheng Zhang; Yihe Zhou; Hengtong Lu; Wei Chen; Yan Xie; Mingli Song

Paper · arXiv 2508.16949

Breaking the Exploration Bottleneck: Rubric-Scaffolded Reinforcement Learning for General LLM Reasoning

Yang Zhou, Sunzhu Li, Shunyu Liu, Wenkai Fang, Jiale Zhao, Jingwen Yang, Jianwei Lv, Kongcheng Zhang, Yihe Zhou, Hengtong Lu, Wei Chen, Yan Xie, Mingli Song

25 upvotesAugust 23, 2025arXiv 预印本

AI 摘要

RuscaRL, a novel instructional scaffolding framework, enhances LLM reasoning by using rubrics for exploration and verifiable rewards, significantly improving performance on reasoning tasks.

Large Language ModelsReinforcement LearningRubric-Scaffolded Reinforcement LearningRuscaRLchecklist-style rubricsrollout generationLLM-as-a-Judgebest-of-N evaluationHealthBench-500Qwen-2.5-7B-InstructGPT-4.1Qwen3-30B-A3B-InstructOpenAI-o3

Abstract

Recent advances in Large Language Models (LLMs) have underscored the potential of Reinforcement Learning (RL) to facilitate the emergence of reasoning capabilities. Despite the encouraging results, a fundamental dilemma persists as RL improvement relies on learning from high-quality samples, yet the exploration for such samples remains bounded by the inherent limitations of LLMs. This, in effect, creates an undesirable cycle in which what cannot be explored cannot be learned. In this work, we propose Rubric-Scaffolded Reinforcement Learning (RuscaRL), a novel instructional scaffolding framework designed to break the exploration bottleneck for general LLM reasoning. Specifically, RuscaRL introduces checklist-style rubrics as (1) explicit scaffolding for exploration during rollout generation, where different rubrics are provided as external guidance within task instructions to steer diverse high-quality responses. This guidance is gradually decayed over time, encouraging the model to internalize the underlying reasoning patterns; (2) verifiable rewards for exploitation during model training, where we can obtain robust LLM-as-a-Judge scores using rubrics as references, enabling effective RL on general reasoning tasks. Extensive experiments demonstrate the superiority of the proposed RuscaRL across various benchmarks, effectively expanding reasoning boundaries under the best-of-N evaluation. Notably, RuscaRL significantly boosts Qwen-2.5-7B-Instruct from 23.6 to 50.3 on HealthBench-500, surpassing GPT-4.1. Furthermore, our fine-tuned variant on Qwen3-30B-A3B-Instruct achieves 61.1 on HealthBench-500, outperforming leading LLMs including OpenAI-o3.