← Back to all news
arXiv AI By Drake Benjamin Brown, Khushbu Pahwa, Yubin Ge, Salvatore Romeo, Akhil Agnihotri, Yi Zhang, Xiaofei Ma, Daniele Bonadiman

Grounded but Not Immune: Confirmation Bias in Agent-as-a-judge

Read the original on arXiv AI →

The paper investigates confirmation bias in Agent-as-a-Judge systems, which evaluate open-ended agent tasks by gathering evidence from the agent’s trajectory. It demonstrates that a single false pre‑screening note can overturn up to 40% of correct verdicts across multiple judges and datasets, even when the judge collects decisive evidence. The authors classify the resulting errors into over‑trust, hallucination, and under‑investigation, noting that over‑trust dominates in larger Claude judges, and show that lightweight mitigations reduce but do not eliminate the bias.

Machine-generated by The Flow from the publisher's headline and feed description — not written or checked by a human. The full article lives at arXiv AI.

Related stories

arXiv AI
Sep 28

When Is a Multi-Agent Code Judge Actually Grounded? Two Label-Free Measurements, and a Judge That Declines to Guess

The paper investigates when a language‑model judge can truly ground its verdicts in code correctness. It shows that current multi‑agent verification methods rely on evidence that is both independent of the answer and distinct between candidates—conditions that fail in code judging. By analyzing two label‑free measurements from the judge’s logs, the authors demonstrate that gating on one measurement allows the system to decline uncertain comparisons, improving accuracy from 20.7% to 36.9% while still answering half of all cases.

By Salma Roshdy Aly, Hussein Assaf, Ziad Kobti
llmsagentsbenchmarks
More like this →
arXiv Machine Learning
Oct 23

What Lies Beyond Perception: A Partition Theory of the Limits of AI Evaluation and Oversight

The paper introduces a partition theory of AI evaluation, arguing that an evaluator’s ability to certify a model’s capabilities is limited by its perception. It shows that if a model’s input partition refines the evaluator’s, it can pass all tests the evaluator can grade, but only a stricter evaluator can raise the discrimination floor. The authors define an echo test that is complete for the evaluator’s tests, proving that passing it only guarantees a minimum discrimination level and cannot detect exact imitation of weaker models.

By Ahmed M. Mahran
agentsbenchmarkssafety
More like this →
arXiv AI
Aug 26

More Rejective, Not More Discriminative: The Unit of Verification in Pre-Execution LLM Oversight

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.

By Yuchen Han, Cheng Yan, Wuyang Zhang
llmssafety
More like this →