一作ACL 2026Poster· Natural Language Processing
Data Efficient RLVR via Off-Policy Influence Guidance
Erle Zhu · Dazhi Jiang · Yuan Wang · Xujun Li · Jiale Cheng · Yuxian Gu · Yilin Niu · Aohan Zeng · Jie Tang · Minlie Huang · Hongning Wang
摘要
Data selection is a critical aspect of Reinforcement Learning with Verifiable Rewards (RLVR) for enhancing the reasoning capabilities of large language models (LLMs). Current data selection methods are largely heuristic-based, lacking theoretical guarantees and generalizability. This work propose...