An Empirical Study of Reward Specification and Benchmark Reliability in GRPO-based LLM Unlearning
Practical LLM unlearning is usually evaluated through two objectives: suppress target-specific knowledge and preserve non-target utility. In generative QA, this leaves a third behavior underspecified: when a target-adjacent prompt admits a broader answer without target-specific leakage, the model should answer at that level rather than leak, evade, or refuse. We study this specification problem in a controlled LoRA-GRPO RWKU setting, comparing four reward designs that span lexical suppression, anti-refusal shaping, rubric-based broad answering, and an explicit refusal contrast, with and without SFT warm-up. The experiments show that optimization success is not equivalent to behavioral unlearning: RWKU forget scores, held-out completion audits, terminal training-rollout audits, and training dynamics can point to different conclusions. We trace these disagreements to reward-hacking endpoints, policy-support limits in GRPO, benchmark probes that miss endpoint changes, and rewards that can select broad-topic answering with low semantic leakage during optimization.
Code (0)
등록된 구현이 없습니다.
Similar Papers 제목 키워드 기반
Jailbreaking as a Reward Misspecification Problem
The widespread adoption of large language models (LLMs) has raised concerns about their safety and reliability, particularly regarding their vulnerability to adversarial attacks. In this paper, we propose a novel perspec…
Red TeamingAverage Reward Reinforcement Learning for Omega-Regular and Mean-Payoff Objectives
Recent advances in reinforcement learning (RL) have renewed focus on the design of reward functions that shape agent behavior. Manually designing reward functions is tedious and error-prone. A principled alternative is t…
Reinforcement Learning (RL)ARM-FM: Automated Reward Machines via Foundation Models for Compositional Reinforcement Learning
Reinforcement learning (RL) algorithms are highly sensitive to reward function specification, which remains a central challenge limiting their broad applicability. We present ARM-FM: Automated Reward Machines via Foundat…
Zero-shot GeneralizationReinforcement LearningModel-Free Reinforcement Learning for Symbolic Automata-encoded Objectives
Reinforcement learning (RL) is a popular approach for robotic path planning in uncertain environments. However, the control policies trained for an RL agent crucially depend on user-defined, state-based reward functions.…
reinforcement-learningReinforcement LearningReinforcement Learning (RL)SmartEval: A Benchmark for Evaluating LLM-Generated Smart Contracts from Natural Language Specifications
We introduce SmartEval, a benchmark for systematically evaluating the quality of Solidity smart contracts generated by large language models (LLMs) from natural language specifications. SmartEval provides a corpus of 9,0…