HarmRLVR: Weaponizing Verifiable Rewards for Harmful LLM Alignment.

Yuexiao Liu, Lijun Li, Xingjun Wang, Jing Shao

Browse the full ACL paper archive.