Exploring the Potential of Offline RL for Reasoning in LLMs: A Preliminary Study
원문 초록
Despite significant advances in long-context reasoning by large language models (LLMs), primarily through Online Reinforcement Learning (RL) methods, these approaches incur substantial computational costs and complexity. In contrast, simpler and more economical Offline RL methods remain underexplored. To address this gap, we investigate the effectiveness of Offline RL methods, specifically Direct Preference Optimization (DPO) and its length-desensitized variant LD-DPO, in enhancing the reasoning capabilities of LLMs. Extensive experiments across multiple reasoning benchmarks demonstrate that these simpler Offline RL methods substantially improve model performance, achieving an average enhancement of 3.3\%, with a particularly notable increase of 10.1\% on the challenging Arena-Hard benchmark. Furthermore, we analyze DPO's sensitivity to output length, emphasizing that increasing reasoning length should align with semantic richness, as indiscriminate lengthening may adversely affect model performance. We provide comprehensive descriptions of our data processing and training methodologies, offering empirical evidence and practical insights for developing more cost-effective Offline RL approaches.
함께 살펴볼 논문
공통 태그를 근거로 연결했습니다. 실제 인용 관계나 논문의 우수성을 뜻하지 않습니다.
Fully Offline Reinforcement Learning
Sparse-Reg: Improving Sample Complexity in Offline Reinforcement Learning using Sparsity
DR-SAC: Distributionally Robust Soft Actor-Critic for Reinforcement Learning under Uncertainty
Offline RL with Smooth OOD Generalization in Convex Hull and its Neighborhood
From Novelty to Imitation: Self-Distilled Rewards for Offline Reinforcement Learning
Robust Bandwidth Estimation for Real-Time Communication with Offline Reinforcement Learning
토론
이 논문과 연결된 포럼 글이 아직 없습니다.