arXiv AI By Yupeng Tang, Mingfeng Lin

When Expert Disagreement Hurts: Prestige and Selectivity in LLM Decision Revision

Read the original on arXiv AI →

The study introduces a two‑pass audit to evaluate how language models revise decisions when faced with disagreement. By fixing the question, evidence, and initial answer, the audit isolates the effect of different revision contexts—neutral reflection, anonymous disagreement, or disagreement from a domain expert—on accuracy. Across five models and 1,989 evidence‑grounded decisions in medicine, science, and law, expert disagreement rarely improves accuracy, often reducing it by up to 16.1 points, and the number of reversals is a poor indicator of harm.

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.

arXiv AI
Sep 30

When Should LLMs Trust Their Own Revisions? A Risk-Aware Study of Intrinsic Self-Correction

The paper investigates intrinsic self‑correction, where a language model revises its own answer without new evidence. Across 29 open‑weight LLMs on BoolQ, GSM8K, and Corr2Cause, the study tracks how revisions change correctness, revealing that while some models improve significantly, others lose a notable fraction of correct answers. The authors compare three runtime strategies—keeping the initial answer, always accepting the revision, and selectively gating revisions—and find that the best approach depends on the model and task, suggesting that self‑correction should be treated as a revision policy rather than a uniformly beneficial second pass.

By Tianzhu Zhang
arXiv AI
Aug 26

Right Diagnoses, Decorative Reasoning:A Perturbation Audit of Medical Chain-of-Thought

The study evaluates whether the chain-of-thought (CoT) rationales produced by medical language models truly influence their answers. Using a 30‑operator perturbation audit that modifies both the question and the CoT (e.g., severity reversal, negation flip, demographic swap, evidence ablation), the authors found that 72.9% of edits did not change the model’s answer—a high Chain‑Decoupling Rate (CDR). Across 14 large language models and four medical QA benchmarks, the CoT text had little impact on accuracy, and removing CoT prompting did not reduce performance. "whyItMatters":"The findings suggest that current medical CoT outputs may be more documentation than genuine reasoning, highlighting the need for better faithfulness checks in clinical AI systems."

By Mengzhu Xu, Jifan Gao, Xia Jiang, Yaoxin Wu, Xi Long
arXiv AI
Oct 26

Validating the Comparison, Not Just the Scorer: Why Overall Accuracy Can Mislead in Interactive LLM Evaluation

The paper argues that relying solely on overall accuracy or F1 scores to validate automatic scorers in LLM evaluation can conceal prompt‑specific errors that distort comparative results. By auditing six models on 180 invented‑entity questions across six prompts, the authors show that a scorer’s overall F1 of .916 masks substantial variation—up to 33.3%—in error rates depending on the prompt. In a second audit, differences in measured prompt effects dropped dramatically when multiple raters were used, highlighting the need to validate the exact comparison rather than the scorer alone.

By Ancil Crayton
arXiv Machine Learning
Sep 25

Return or Revise? Learning When Revision Helps Retrieval-Augmented QA

The paper investigates when it is better to return an existing draft answer or revise it using retrieved evidence in retrieval‑augmented QA systems. By grading both the draft and its candidate revision with the same correctness judge, the authors define a paired effect called recoverability and train policies to predict it before revision. Experiments on 25,870 open‑domain questions show that a recoverability‑based scorer outperforms a draft‑correctness scorer across multiple Llama setups, improving accuracy–revision trade‑offs and closing a significant portion of the oracle gap, though it still applies harmful revisions in a substantial fraction of cases.

By Nicholas Kashani Motlagh, Tim Anderson, Jeremy Gwinnup, Grant Erdmann