User Feedback Provides a Unique Signal that LLMs Can not Detect

Harnessing naturally occurring feedback from user interactions offers a promising learning signal for Large Language Models (LLMs). However, recent studies suggest this feedback is inherently noisy and difficult to leverage effectively. We challenge this conception by demonstrating that user feedback is a highly actionable signal for improvement, and that its perceived ineffectiveness stems from a systematic bias in current evaluation paradigms. To isolate the usefulness of feedback, we construct synthetic data with a definitive ground truth, alongside naturalistic data to validate that our findings hold in real-world scenarios. By comparing model revisions generated with and without access to feedback across both settings, we show that feedback-informed revisions resolve targeted issues at significantly higher rates than baseline revisions. Finally, we expose the root of the evaluation bias: when a model successfully fixes an issue exclusively due to feedback, LLM judges frequently fail to identify the genuinely corrected response, systematically preferring inferior baseline outputs instead.

Anchoring Bias in LLM-as-a-Judge Systems: Prior Scores Compromise Evaluation Independence

Large language models (LLMs) increasingly assess generated content, giving rise to the LLM-as-a-Judge paradigm. These systems now score outputs, filter content, and gate iterative refinement in production pipelines, where each judgment is often assumed to be independent of earlier evaluations. We test this assumption using three prompt conditions: no metadata, revision framing, and anchored metadata containing revision, attempt, and prior-score fields. We show that prior scores, even when included only as context metadata, anchor judgments and systematically shift ratings toward their values. Across 192,000 attempted evaluations (185,271 successful), seven out of the eight evaluated models have 95% task-stratified bootstrap intervals below zero for the total anchored-metadata effect on 20 fixed texts. Cohen's $d$, a standardized measure of the difference between score distributions, reaches an absolute value of 0.71. Token-level analysis of selected model-task probes suggests a threshold-like response pattern: introducing anchored metadata produces a marked redistribution of output-score probabilities, while changing the anchor value within the tested below-threshold range produces comparatively little additional variation. On categorical industry data with human-labeled ground truth, anchored metadata blocks 48% of error corrections and flips 10.18% of correct judgments toward an assigned wrong label, demonstrating the bias extends beyond numerical scoring to categorical decisions. Neither Chain-of-Thought nor a metadata-disregard warning reduces the total effect, although the warning improves the paired accuracy effect relative to baseline in the industry experiment. Reliable LLM evaluation demands careful context engineering rather than an assumption of impartiality. Effective mitigation must be validated for the intended model and task or domain.

Style Wins, Substance Loses: A Diagnosis of LLM-as-Judge in Idea Generation

However, whether these judges truly evaluate the scientific substance of ideas or are influenced by superficial stylistic presentation remains an open question. To address this question, we propose SciStyleBench, a unified three-component benchmark for diagnosing and mitigating stylistic bias in LLM-based idea evaluation: (i) First, SciStyleStage, a three-stage evaluation environment that applies controlled stylistic perturbations to fixed scientific content across three settings no context, fixed-domain context, and open-domain retrieval context, covering 600 scientific ideas and 15 style variants, with 9,000 evaluation instances per setting; (ii) Second, SciStyleMetrics, a set of quantitative measures, including Style Bias Index (SBI), Substance Recognition Rate (SRR), and Adversarial Win Rate (AWR), to characterize how stylistic variation affects scoring stability, substance discrimination, and ranking robustness; (iii) Third, SciStyleExtractor, a plug-and-play evaluation module that separates presentation style from scientific content by predicting style type and deviation before style-conditioned evaluation, enabling us to assess whether style awareness reduces stylistic bias. Experiments on SciStyleBench show that direct LLM judges remain sensitive to writing style and struggle to distinguish scientific substance. In contrast, SciStyleExtractor reduces SBI from 0.566 to 0.501 while increasing SRR and AWR from 0.504 and 0.554 to 0.759 and 0.899, respectively. These results suggest that robust idea evaluation requires invariance to stylistic variation without sacrificing sensitivity to scientific substance. Overall, SciStyleBench provides a systematic framework for identifying, quantifying, and mitigating stylistic bias in scientific idea evaluation.

Self-Generated Text Recognition: Quality Heuristics, Cross-Task Transfer, and Downstream Bias in LLM Evaluation

Self-Generated Text Recognition (SGTR)--the ability of an LLM to identify its own outputs--poses risks to AI safeguards that rely on LLMs as evaluators or monitors: an LLM may recognize outputs from other copies of the same model and make biased judgments or collude outright. Prior work has drawn conflicting conclusions about whether current models possess significant SGTR capabilities. We explain these disagreements by identifying key experimental design choices--which we term operationalizations--that drive divergent results. Evaluating 13-21 models across six presentation operationalizations and four task-domain operationalizations, we find that accuracy varies substantially with evaluation format (pairwise vs individual assessments of text), conversation format (presenting candidate text in user tags vs assistant tags), and the domain of the task used to generate candidate text (e.g., coding vs summarization). We corroborate previous observations that a quality heuristic--models attributing authorship to text they perceive as higher quality--is a dominant confound. We also find that improving a model's SGTR performance via supervised fine-tuning (SFT) on one operationalization can generalize to others, and can increase the model's preference for its own outputs when it acts as a judge in the AlpacaEval framework. Our results suggest that, despite confounds, some models possess practical SGTR capabilities, and that SGTR should be monitored and considered in the design of safety-critical AI applications.

Self- and Other-Labels Induce Bidirectional Bias in LLM Judges

As LLM-as-a-judge systems become increasingly widespread, self-preference in LLMs -- the tendency to favor one's own outputs -- raises growing concerns about evaluation reliability. However, it has been studied predominantly on generated text, where stylistic features and response quality are inevitably conflated. As a result, existing measurements cannot separate genuine self-preference from these confounds. We address this by changing the object of evaluation: instead of judging generated text, ten LLMs assess narrative constraint selections, which carry no model-specific stylistic fingerprint yet retain a recoverable model-specific signature. We run two experiments that yield distinct findings. Under blind evaluation, self-preference largely disappears once selection quality and evaluator severity are controlled. It vanishes on three of four rubric dimensions and reverses on the fourth, where judges rate their own selections as less original. Under matched quality, however, self- and other-labels alone -- without naming any model -- shift scores bidirectionally: LLM judges inflate scores for self-labeled selections and deflate those for other-labeled ones regardless of the selection's actual source. We make two contributions: 1) authorship attribution is a distinct driver of evaluation bias, and 2) open-ended, ground-truth-free tasks can serve as controlled instruments for studying LLM judge behavior.

Bias in Filter Feature Selection Evaluation: A Meta-Analysis of Datasets, Baselines, and Experimental Design Choices

Background: Since 1990 many feature selection methods have been proposed across heterogeneous applications. To validate the usefulness of a new method, it needs to be compared against at least one baseline method from the existing literature on a feature selection task using at least one dataset. Recent developments in tabular Deep Learning (DL) and data valuation in Machine Learning (ML) suggest that the evaluation of new methods, algorithms, and models may be consciously or unconsciously biased. We hypothesise that a similar trend exists in feature selection (FS), particularly in filter feature selection (FFS). The aim of this study is therefore to examine FFS studies to identify factors that influence the evaluation and that might consist entry point for biases in order to recommend stronger principles for FFS evaluation. Methods: We analyse a sample of 28 high profile FFS studies published between 1994 and 2025. The analysis provides reflections on how to examine FFS studies, highlights lessons learned throughout the process, and gives five evidence-based recommendations for future FFS evaluation. Results: Multivariate Linear Regression analysis achieved a score of $R^2=0.33$. It means that 33% of the variance in the performance of new methods against chosen baselines (win rate) is explained by the number of datasets (#Datasets), the number of baselines (#Baselines), and the number of new methods (#NewMethods). Discussion: $R^2=0.33$ is considered medium explanation; which is promising given that this is the first such study. The medium explanation result is due to the fact that win rate is influenced by additional factors such as the maturity of the feature selection domain, the type of datasets and baselines, and the simplicity of the regression model used to explain the relationship.