一作ACL 2026Poster· Natural Language Processing
H arm RLVR : Weaponizing Verifiable Rewards for Harmful LLM Alignment
Yuexiao Liu · Lijun Li · Xingjun Wang · Jing Shao
摘要
Recent advancements in Reinforcement Learning with Verifiable Rewards (RLVR) have gained significant attention due to their objective and verifiable reward signals, demonstrating strong performance in reasoning and code generation tasks. However, the potential safety risks associated with RLVR re...