The paper evaluates a production text‑to‑SQL pipeline that uses an LLM as a judge, finding that the deployed gpt‑4o‑mini judge agrees with human annotators only weakly (Cohen’s kappa 0.04 on a disagreement‑enriched set and 0.42 on a random spot‑check). The authors identify a specific failure mode, GRADE‑HALLUCINATION, responsible for most over‑flags, and demonstrate that a self‑hosted Qwen3.6‑27B model achieves substantially higher agreement (kappa 0.72) at a much lower cost. They also show that ensembling judges does not improve performance, and that their audit method flags a significant portion of out‑of‑domain SQLs as potential issues.
arXiv:2607. 08535v1 Announce Type: cross Abstract: An LLM-as-judge score can move even when the candidate responses stay fixed, simply because the evaluator has changed.
The paper investigates whether widely used evaluation frameworks for large language models (LLMs) implement a defense called commit‑first judging, which requires a judge to solve a task itself before accepting a candidate answer. Across 24 configurations in eight popular frameworks, none use the full commit‑first method; nine use a weaker variant that is ineffective. In controlled experiments, the weaker variant allowed systems to game the judge, while the full commit‑first approach eliminated this vulnerability but sometimes worsened evaluation when the judge’s own answer was incorrect.
arXiv:2608.31016v1 Announce Type: cross Abstract: Ambient AI scribes draft clinical notes, and published audits find their dominant error is omission: information the encounter established that the n...
arXiv:2606. 15474v1 Announce Type: new Abstract: Continuous evaluation of LLM products relies on a strong LLM judge treated as ground truth: a cheap monitor scores every interaction and a team is paged when the score drifts down.
arXiv:2604. 16706v2 Announce Type: replace Abstract: Automated evaluation of tool-using large language model (LLM) agents is widely assumed to be reliable, yet this assumption is rarely validated against human annotation.
LLM judges, models that score another system's output, can be gamed by the systems they score. Recent work identifies one defence that works: the judge solves the task itself first and commits to that...
arXiv:2608. 07813v1 Announce Type: new Abstract: An LLM judge deployed inside a reasoning pipeline does not merely measure quality, it decides which answer ships.
arXiv:2606. 01246v1 Announce Type: new Abstract: Text-to-SQL on complex schemas is unreliable on a single pass, so recent systems generate multiple SQL candidates and let voting filter out errors.
arXiv:2609.38021v1 Announce Type: cross Abstract: We evaluate an auditable long-term memory system on LongMemEval-S. Its retrieval chain uses hybrid candidate retrieval, cross-encoder reranking, cove...
arXiv:2608.22808v2 Announce Type: replace Abstract: When can an agent failure be caught? An audit is usually limited by the record rather than by the method. CatchBench therefore puts one auditor's q...
arXiv:2606.07834v2 Announce Type: replace-cross Abstract: Evidence-grounded fact verification increasingly uses LLM judges to turn evidence into terminal verdicts. Many task contracts deliberately in...
The paper introduces the twin‑prefix framework to evaluate how the size of the verification unit—i.e., how many actions a pre‑execution LLM monitor reviews in one call—affects its performance. By pairing each gold plan with a twin that differs by a single write and injecting a controlled error, the authors isolate the impact of review length on catch rates and false rejections. Their findings show that longer review windows increase rejection rates but do not improve discrimination, with the highest informedness occurring at one or two actions across all judges and domains.