The paper investigates how different reward specifications affect the reliability of unlearning in large language models using a LoRA-GRPO framework. It compares four reward designs—lexical suppression, anti-refusal shaping, rubric-based broad answering, and explicit refusal contrast—both with and without a supervised fine-tuning warm-up. The results reveal that successful optimization does not guarantee behavioral unlearning, as various evaluation metrics can yield conflicting conclusions due to reward-hacking, policy-support limits, and benchmark probe limitations.
arXiv:2608. 15445v1 Announce Type: new Abstract: When a reward is correct on every training example yet consistent with more than one goal, a model can acquire an unintended one, a failure known as goal misgeneralization.
arXiv:2608. 07531v1 Announce Type: cross Abstract: Search-augmented language agents should retrieve external information only when necessary and ground their answers in retrieved evidence.
arXiv:2605.20066v2 Announce Type: replace Abstract: Knowledge graph question answering seeks to translate natural language questions into executable queries over knowledge graphs, but existing approa...
arXiv:2608.30005v1 Announce Type: new Abstract: Rubric-based reinforcement learning extends RL beyond tasks with exact answers or rule-based verifiers by scoring responses against instance-specific c...
arXiv:2509. 25760v2 Announce Type: replace-cross Abstract: While large language models (LLMs) have demonstrated strong performance on factoid question answering, they are still prone to hallucination and untruthful responses, particularly when tasks demand information outside their parametric knowledge.
arXiv:2609.39321v1 Announce Type: cross Abstract: Group Relative Policy Optimization (GRPO) has emerged as a memory-efficient reinforcement fine-tuning (RFT) technique for reasoning-intensive tasks....
arXiv:2609.24290v1 Announce Type: new Abstract: Instruction-tuned LLMs faced with underspecified queries often commit to a single interpretation rather than ask for clarification, producing confident...
arXiv:2608. 11922v2 Announce Type: replace-cross Abstract: Predictive-distribution entropy is a strong answer-selection rule in retrieval-augmented generation (RAG) for question answering: across five QA benchmarks, selecting the answer a frozen respondent LLM produces with the lowest answer-token entropy lifts mean $F_1$ from 0.
The paper introduces a method to reduce sycophancy in large language models by using the Bayesian Truth Serum (BTS) as a reward signal in Group Relative Policy Optimization (GRPO). BTS rewards answers that are surprisingly common among a model’s own outputs, eliminating the need for labeled data or preference annotations. Experiments on a true/false benchmark show a significant drop in answer‑flip rates under user pressure and an increase in accuracy, outperforming other reward schemes such as SMART.
arXiv:2606. 27291v1 Announce Type: new Abstract: Job-search platforms rely on low-bandwidth query interfaces that often fail to capture the high-dimensional complexity of candidate profiles.
The study investigates whether language models can explicitly report constraints they have learned through post‑training fine‑tuning. Using constrained recipe generation with five banned ingredients, the authors compare supervised fine‑tuning (SFT) and Group Relative Policy Optimization (GRPO) against an untrained baseline on a Constraint Awareness Benchmark. Both fine‑tuning methods increase behavioral compliance from 4% to about 90% but reduce explicit constraint reporting and erode retained third‑person knowledge, with GRPO showing more destructive effects. The results suggest that reward‑based signals may suppress constraints context‑independently, and that models fail to enumerate constraints on request even when they can avoid them internally.
arXiv:2607. 20090v1 Announce Type: cross Abstract: Retrieval-augmented large language models frequently face contexts that interleave useful evidence with misleading statements or instruction-like content.