arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.17684cs.LGcs.AI
收藏

CodeScaler: 通过奖励模型扩展代码大语言模型的训练和测试时间推理

CodeScaler: Scaling Code LLM Training and Test-Time Inference via Reward Models

LARK, HKUST(GZ)(LARK,香港科技大学(广州));Kuaishou Technology(快手科技);UCL(伦敦大学学院);UZH(苏黎世联邦理工学院);NUS(国立新加坡大学)

Xiao Zhu, Xinyu Zhou, Boyu Zhu, Hanxu Hu, Mingzhe Du, Haotian Zhang, Huiming Wang, Zhijiang Guo

AI总结:

本文提出CodeScaler,一种通过奖励模型扩展代码生成模型的训练和测试时间推理的框架,通过精心编纂的偏好数据和语法感知的代码提取,实现了在四个编码基准上比基于执行的RL提升1.55分,在Qwen3-14B-Base上提升4.23分,并在无测试用例的情况下通过合成数据进一步提升14.64分,同时在推理时间减少10倍的延迟,且在代码、通用和推理领域均优于现有奖励模型。

摘要:

基于可验证奖励的强化学习(RLVR)通过利用单元测试的执行反馈推动了代码大语言模型的最新进展,但其可扩展性从根本上受到高质量测试用例可用性和可靠性的影响。我们提出CodeScaler,一种奖励模型,旨在扩展代码生成的强化学习训练和测试时间推理。CodeScaler是在经过验证的代码问题上精心编纂的偏好数据上训练的,并结合语法感知的代码提取和保持有效性的奖励塑造,以确保稳定和稳健的优化。在四个编码基准上,CodeScaler在Qwen3-8B-Base上比基于执行的RL提升1.55分,在Qwen3-14B-Base上提升4.23分。通过进一步扩展到44K问题并添加额外的合成数据,CodeScaler在无任何测试用例的情况下,相对于基础模型提升了14.64分。在推理时间,CodeScaler作为有效的测试时间扩展方法,实现了与单元测试方法相当的性能,同时在推理时间减少了10倍的延迟。此外,CodeScaler在RM-Bench上不仅在代码领域(+3.3分)上优于现有奖励模型,还在通用和推理领域(平均+2.7分)上也表现优异。

英文摘要:

Reinforcement Learning from Verifiable Rewards (RLVR) has driven recent progress in code large language models by leveraging execution-based feedback from unit tests, but its scalability is fundamentally constrained by the availability and reliability of high-quality test cases. We propose CodeScaler, a reward model designed to scale both reinforcement learning training and test-time inference for code generation. CodeScaler is trained on carefully curated preference data derived from verified code problems and incorporates syntax-aware code extraction and validity-preserving reward shaping to ensure stable and robust optimization. Across four coding benchmarks, CodeScaler consistently outperforms execution-based RL by +1.55 points on Qwen3-8B-Base and +4.23 points on Qwen3-14B-Base. By further scaling to 44K problems with additional synthetic data, CodeScaler yields +14.64 points improvement over the base model without requiring any test cases. At inference time, CodeScaler serves as an effective test-time scaling method, achieving performance comparable to unit test approaches while providing a 10-fold reduction in latency. Moreover, CodeScaler surpasses existing reward models on RM-Bench not only in the code domain (+3.3 points), but also in general and reasoning domains (+2.7 points on average).

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑