探索全球前沿学术脉络

Are LLM Evaluators Really Narcissists? Sanity Checking Self-Preference Evaluations

Recent research has shown that large language models (LLMs) favor their own outputs when acting as judges, undermining the integrity of automated post-training and evaluation workflows. However, it is difficult to disentangle which behaviors are explained by narcissism versus experimental confounds. Specifically, LLM evaluators may deliver self-preferring verdicts when comparing responses to questions they fail on; these verdicts may not depend on the identity of the author, but on evaluator quality. We correct this by directly comparing the judge's voting distribution in cases where it evaluates itself versus another model. This evaluator quality baseline reveals that only 51% of examples in previous findings retain statistical significance against this null hypothesis, covering 89.6% of total self-preference probability mass. Finally, we compare the entropy of voting distributions, suggesting uncertainty-driven overlap, and show that our procedure enables more careful documentation against the backdrop of judge-bias research.

Dynestyx: A Probabilistic Programming Library for Dynamical Systems

arXiv:2606.16985v1 Announce Type: cross Abstract: State-space models (SSMs) are the standard formalism for Bayesian treatment of dynamical systems, with natural applications in statistics, signal processing, and machine learning. Despite their importance in both theory and application, dynamical systems have proven difficult to incorporate in modern probabilistic programming languages (PPLs), making state-of-the-art methods less accessible to practitioners and introducing friction in following the "Bayesian workflow." We introduce dynestyx, a probabilistic programming library with first-class support for SSMs, including state-of-the-art methods in the estimation of both states and parameters. Through a single, unified interface, users may specify arbitrary priors for discrete-time or continuous-time dynamical systems, perform inference over mixed-effect data, and make state and parameter estimates with principled uncertainty quantification.

Breaking the Mirror: Activation-Based Mitigation of Self-Preference in LLM Evaluators

Large language models (LLMs) increasingly serve as automated evaluators, yet they suffer from "self-preference bias": a tendency to favor their own outputs over those of other models. This bias undermines fairness and reliability in evaluation pipelines, particularly for tasks like preference tuning and model routing. We investigate whether lightweight steering vectors can mitigate this problem at inference time without retraining. We introduce a curated dataset that distinguishes self-preference bias into justified examples of self-preference and unjustified examples of self-preference, and we construct steering vectors using two methods: Contrastive Activation Addition (CAA) and an optimization-based approach. Our results show that steering vectors can reduce unjustified self-preference bias by up to 97\%, substantially outperforming prompting and direct preference optimization baselines. Yet steering vectors are unstable on legitimate self-preference and unbiased agreement, implying self-preference spans multiple or nonlinear directions. This underscores both their promise and limits as safeguards for LLM-as-judges and motivates more robust interventions.

Long-term Penetrance of Disease Variants in Genes Prioritized for Genomic Newborn Screening: Evidence from Adult Biobanks

Importance: Genomic newborn screening (gNBS) is a potential public health intervention, but its positive predictive value (PPV) remains uncertain. Estimating the prevalence and penetrance of pathogenic and likely pathogenic (P/LP) variants in genes prioritized for screening may clarify the long-term PPV and clinical utility of gNBS. Objective: To compare ICD-based ascertainment, electronic medical record (EMR) review, and clinical assessment of genetic disorders in adults with P/LP variants in 54 genes prioritized for gNBS. Design: Two-cohort observational study with EMR review and clinical assessment in the hospital-based cohort. Setting: The U.K. Biobank (UKB) and Mass General Brigham Biobank (MGBB). Participants: 451,877 adults from the UKB and 53,371 from the MGBB, all with exome sequencing data. Exposures: P/LP variants in 54 genes prioritized through expert consensus for gNBS, in genotypes consistent with each gene's inheritance pattern. Main outcomes and measures: The primary outcome was the absolute difference in the proportion of MGBB participants identified as affected by ICD versus EMR ascertainment. Secondary outcomes included findings from clinical assessments of undiagnosed MGBB participants, corrected UKB penetrance estimates, and extrapolation to U.S.. annual birth cohorts and living adults. Results: P/LP variants were identified in 665 UKB participants (0.15%) and 82 MGBB participants (0.15%), approximately 1 in 650. In MGBB, EMR review revealed that 58/82 individuals (70.7%) were undiagnosed, although 25 of 58 (43.1%) had documented symptoms. Disease-associated ICD codes were found in 39.0% (32/82) of participants, whereas EMR review identified symptoms in 59.8% (49/82, McNemar P