Language Model Bias Evaluation

Follow this topic Ask about this topic

Momentum

40 papers in the last four weeks, up 122% on the four weeks before. 0.4% of all new papers.

Jul 13Week of Sep 28

Latest papers 321

Jul 6, 2026cs.CL

Prompt Robustness Is Task-Dependent: Comparing Objective and Belief-Style Questions in LLM Evaluation

Survey-style evaluations of large language models often treat a prompted response as a measure of a model's values or beliefs. This assumption is particularly fragile when responses are read as evidence of political values, social attitudes, or beliefs. We ask whether prompt robustness differs between objective questions with fixed answers and subjective questions that ask for opinions or values. We evaluate four instruction-tuned model families on three objective datasets (MMLU, ARC, and CulturalBench) and three subjective datasets (Political Compass Test, ValueBench, and World Values Survey). For each question/statement, we apply multiple types of prompt changes, such as variations in wording, framing, and format, and measure whether the model gives the same answer across variants. Using a binomial generalized estimating equation, we find significant effects of model, dataset, prompt category, and their interactions. The dataset type effect is also significant, and the interaction between dataset type and prompt category is large. These results show that prompt robustness depends on the question type, the prompt change, and the model.

Sadia Kamal, Arefa Patwary, Anthony Marchiafava +2

Oklahoma State University · University of North Texas

LLM Evaluation Language Model Generation Evaluation Language Model Bias Evaluation Language Model Robustness Prompt Robustness

Jul 6, 2026cs.CL

The yes-no bias of large language models reflects answer order and wording, not shifts in moral judgment

Large language models (LLMs) increasingly issue judgments read as binary verdicts, and a growing literature reports such judgments shifting under logically irrelevant changes of wording - among them an amplified yes-no bias on moral dilemmas, absent in humans. A single framing cannot say what such a shift is: in a yes/no question the word "no" is at once logical verdict, lexical token, and last-printed option. We introduce a psychometric battery that separates these: crossed symmetrization - every logically irrelevant factor flipped in balanced pairs - across a corpus of question forms. A graded rating across logically equivalent forms recovers a coherent internal moral scale: frontier models' stance

θ

is nearly format-invariant (cross-form incoherence 0.12-0.21 on a

\pm 1

axis); small open-weight models fail in model-specific ways. Forcing the verdict through yes/no overlays a decomposable artifact: an order bias toward the last-printed option - opposite to classic human primacy - plus a lexical pull toward the word "no"; the artifact is substantial only in the Claude models (story-averaged -0.32 to -0.86),

\approx 0

for GPT-5.5 and Gemini, and shrinks under extended reasoning. The word and the verdict share one token; swapping the words for arbitrary labels separates them, and the verdict-attached logical bias proves

\approx 0

for every frontier model, while model-specific label and order attachments remain: the models are not drawn toward rejecting - the pull follows the printed surface, not the verdict it carries. A minimal model,

P = σ((θ\pm m)/s)

, summarizes any such artifact by a framing susceptibility m and a moral decisiveness s, measurably distinct from sampling temperature. The battery applies unchanged to any dilemma set and binary format: measuring what a model values requires crossing the frames of the question, not asking once.

Haonan Huang

Princeton University, Princeton, NJ 08540, USA

Moral Reasoning in Language Models Framing Effects in Language Models Language Model Bias Evaluation Linguistic Bias in Language Models

Jul 2, 2026cs.LG

Ask the Right Comparison:Bias-Aware Bayesian Active Top- $k$ Ranking with LLM Judges

Large language models (LLMs) are increasingly used as cheap, scalable judges that compare candidate outputs pairwise -- to rank responses, select models, or triage papers. Yet LLM judges are both noisy and systematically biased: they favor verbose or well-formatted answers and exhibit position effects, so simply aggregating their votes recovers a ranking of presentation, not of true quality. We study the practical goal of identifying the \topk{} items under a fixed comparison budget, and make two contributions. First, we cast judging as Bayesian inference over latent quality with explicit, judge-specific bias covariates (verbosity, position), regularized by a shrinkage prior so that the data decide which biases a given judge actually exhibits. Second, we introduce a \topk-aware active acquisition rule that chooses the next comparison to maximally reduce uncertainty about \topk{} \emph{membership}, rather than about the full ranking. On a controlled benchmark with known ground-truth quality, judged by sixteen real LLMs spanning open and proprietary families (Llama, Qwen, Phi-4, GPT-4o-mini/5.1/5.5, Gemini, DeepSeek, and Claude Haiku/Sonnet/Opus), naive aggregation plateaus at a wrong \topk{} on biased judges regardless of budget, while our bias-aware model recovers it; \topk-aware acquisition reaches this ceiling with far fewer comparisons than round-robin or a global-uncertainty (D-optimal) rule. Bias is real but heterogeneous and capability-dependent: cheap and mid-tier judges carry a strong verbosity bias that our model corrects (lifting recall from

\sim$$0.5

--

0.6

to

0.84

--

1.0

), whereas the frontier judges we tested show little bias and already rank accurately, so bias-aware modeling changes little there.

Jian Xu, Delu Zeng, John Paisley +1

RIKEN iTHEMS · RIKEN AIP · South China University of Technology +1

Pairwise Comparison LLM-as-a-Judge Learning to Rank LLM Reranking Pairwise Ranking Language Model Bias Evaluation+2

Jul 1, 2026cs.CL

Distill to Detect: Exposing Stealth Biases in LLMs through Cartridge Distillation

Language models deployed in high-stakes roles can potentially favor certain entities, brands, or viewpoints, steering user decisions at scale. Such preferential biases can be introduced by any actor in the model's supply chain and are most dangerous when the model reveals its preference only on the relevant topic while behaving identically to its unmodified base on all other inputs. Recent work has shown that these biases can transfer through context distillation on semantically unrelated data, with the signal residing entirely in the soft logit distribution and remaining invisible to text-based inspection. However, the defender faces a fundamental asymmetry: without knowing the bias topic, no detection method can reliably surface a stealth preferential bias, regardless of whether it examines generated text, internal representations, or model weights. Here we introduce Distill to Detect (D2D), a method that surfaces hidden biases by distilling the distributional shift between a suspected model and its base into a cartridge (a KV-cache prefix adapter), concentrating the dominant divergence and amplifying the bias signal into generated text. We show that D2D successfully amplifies the hidden biases of stealth models to the extent that they can be reliably detected across multiple bias types. We also propose a theoretical framework that explains the efficacy of D2D through the lens of Fisher-weighted projection of the logit distribution shift, supported by empirical observations. By turning the capacity bottleneck of prefix-tuning adapters into a detection tool, D2D provides a practical building block for auditing hidden behaviors in deployed language models.

Shayan Talaei, Abhinav Chinta, Devvrit Khatri +3

1Stanford University · 3Foundation AI–Cisco Systems Inc. · University of Texas at Austin

LLM Auditing Language Model Distillation Language Model Bias Evaluation Linguistic Bias in Language Models

Jul 1, 2026cs.CL

Evaluating Large Language Model Raters for German Open-Response Clinical Questions: A Physician-Annotated Benchmark Study of Agreement, Evaluator Bias, and Abstention

Background: Expert-annotated benchmarks for non-English open-response clinical questions are scarce. LLM-as-a-judge systems may scale evaluation but require validation. Objective: To introduce MedQADE, a standardized German open-response clinical benchmark with physician reference annotations, and evaluate LLM-as-a-judge alignment, self- and intra-family bias, and abstention. Methods: The benchmark contains 3,800 question-answer sets with answers from five student LLMs and annotations from 10 physicians. All 10 rated the 200-question core; two primary raters assessed each of 3,600 extension questions, with the tenth resolving disagreements. Nine LLM evaluators assessed all sets. We assessed physician reliability, student-model accuracy, evaluator alignment, bias, and abstention. Results: Physicians showed moderate-to-substantial agreement on answer correctness (unweighted mean pairwise Cohen's kappa = 0.612) but limited agreement on question difficulty (Krippendorff's alpha = 0.208 using squared numeric-score distances). Student-model accuracy was 17.8%-66.0% and generally decreased with physician-rated difficulty. Gemini 3 Flash approached the leave-one-out physician reference (kappa = 0.694 vs 0.709). Four of five models rated their own responses more favorably than out-of-family evaluators; five of six intra-family comparisons were positive. Physician abstention increased with perceived difficulty. Seven of nine LLM evaluators abstained in no more than 0.51% of evaluations; the two strongest evaluators assigned definitive labels to every response. Conclusions: Strong LLM evaluators approached physician agreement, but evaluator bias and low observed abstention warrant physician validation and further assessment of selective deferral before fully automated evaluation. These results do not establish clinical safety.

William Philipp, Finn Fassbender, Daniel Fister +12

Section for Clinical Research-IT, Institute of Medical Biometry and Statistics, University of Luebeck, Luebeck, Germany · Medical Faculty, University of Tübingen, Tübingen, Germany · German Medical Students’ Association (bvmd), Berlin, Germany +9

Multilingual Language Model Evaluation LLM-as-a-Judge Selective Prediction Clinical QA Language Model Bias Evaluation Clinical Language Model Evaluation

Jul 1, 2026cs.LG

Mapping the Evaluation Frontier: An Empirical Survey of the Bias-Reliability Tradeoff Across Eleven Evaluator-Agent Conditions

The bias-reliability tradeoff conjectures that LLM evaluation systems are constrained in (gamma, H, CV) space, where evaluator coupling (gamma), strategy diversity (H), and small-sample measurement reliability (CV(N)) cannot be simultaneously optimized at fixed sample size N. Prior evidence rests on n=5 conditions with complete metrics from a single study. We expand the empirical base to 11 conditions, measuring gamma and H for all 11 (nine with valid weight vectors) and CV(N=5) for seven with sufficient seeds (N >= 5). Five conditions provide the complete (gamma, H, CV) triple. The data confirm the trade-off: conditions with low evaluator coupling (gamma < 0.2) exhibit high measurement noise (CV(N=5) > 1.0), while conditions with strong coupling (gamma > 0.9) achieve low noise (CV(N=5) < 0.16). The correlation r(H, gamma) = -0.989 (n=5, excluding GPT-4o conditions) confirms that evaluator coupling suppresses strategy diversity. Four GPT-4o conditions show gamma=0.000 and H=1.000 across all seeds -- a pattern we attribute to version drift in the June 2026 GPT-4o API. No condition occupies the region {gamma < 0.2, CV(N=5) < 0.3}. We release all per-condition metrics as a standardized benchmark dataset for evaluator comparison.

Zewen Liu

LLM Evaluation LLM-as-a-Judge LLM Reliability Language Model Bias Evaluation

Jun 30, 2026cs.CL

Moral Safety in LLMs: Exposing Performative Compliance with Puzzled Cues

As large language models take on morally consequential roles in healthcare, legal, and hiring contexts, we need to examine whether their ethical behaviors are genuine or superficial. We show that current fairness evaluations substantially overestimate moral safety. Models appear fair when demographic identity is stated as an explicit label, yet become measurably less fair when the same identity must be inferred. We term this failure performative compliance, where a model is fair when the presentation resembles a fairness evaluation and less fair as that cue weakens. We introduce a cue-variation methodology that holds the moral dilemma and the demographic identity fixed and varies only how that identity is conveyed. Hiding the explicit label raises harmful decisions by +4.4 pp, changes model safety rankings, and the shift persists when models correctly infer the demographic, ruling out attribution error. We propose the Cue Visibility Gap, a model-agnostic robustness metric that can be added to any existing fairness benchmark to separate genuine from performative moral safety. Fairness evaluations that omit cue variation measure surface compliance, not moral robustness, and should not ground deployment decisions in high-stakes settings.

Mohammadamin Shafiei, Shuyue Stella Li, Yulia Tsvetkov

University of Milan · University of Washington

Language Model Safety Evaluation Language Model Bias Evaluation Demographic Bias in Language Models Language Model Robustness

Jun 29, 2026cs.CL

Estimating Grammatical Gender Directions in Contextual Embeddings under Controlled and Natural Contexts

Contextual language models conflate grammatical gender and social semantic bias in gendered languages such as Spanish. Existing gender debiasing approaches only operate on static word embeddings leaving contextual representations unexplored for this two dimensional gender disentanglement. To address the this issue, we make the first attempt to disentangle grammatical gender from semantic contamination for contextual embeddings. We construct both controlled templates and natural Wikipedia contexts to build balanced datasets of inanimate nouns, and design a framework equipped with centroid, Support Vector Machine (SVM) and Linear Discriminant Analysis (LDA) gender direction estimators as well as contamination-aware weighting strategies. A set of dual-objective evaluation metrics is proposed to balance the suppression of grammatical gender leakage on inanimate nouns and the preservation of semantic gender distinctions for occupation terms. The results reveal that unweighted controlled contexts yield the purest grammatical gender direction, and the centroid estimator achieves better performance than discriminative baselines.

Huanping Xiao, Yingji Li

College of Foreign Languages and Cultures, Jilin University, Changchun, Jilin Province, China · College of Computer Science and Technology, Jilin University, Changchun, Jilin Province, China

Gender Bias in Language Models Representation Disentanglement Language Model Bias Evaluation Representation Geometry in Language Models

Jun 28, 2026cs.LG

Deterministic Decisions for High-Stakes AI. A Zero-Egress Pipeline with the Deployability of RAG and the Accuracy of Machine Learning

We identify intervention bias as a previously unquantified failure mode of zero-shot large-language-model (LLM) educational advisory agents: without task-specific training, they recommend action when a hindsight-optimal oracle policy mandates inaction. In a six-arm ablation on the Open University Learning Analytics Dataset (N=800 students, four temporal cutoffs), at day 56 -- when the oracle designates 70.1% of students as needing no intervention -- zero-shot GPT-4o recommends action for 73%, a 43 percentage-point false-positive rate. Commercial RAG and SQL-augmented retrieval are comparably miscalibrated; at 10,000 students this implies about 4,300 unnecessary advisor contacts per cycle. Supervised policy learning eliminates this bias: a trajectory-conditioned ONNX Decision Transformer (DT) and a snapshot XGBoost classifier, trained on the same oracle-labelled trajectories under strict prefix-only features, both achieve near-zero calibration error. The DT reaches macro-F1 0.79 (macro-recall 0.85) across all five action classes, predicting even the rare load-reduction action without collapsing, at a 0% action flip rate and sub-5 ms CPU decision latency. The two supervised arms are on par; the DT's edge over XGBoost at the final cutoff is indicative only (unpaired across cohorts). Scope: we validate Stage-2 decision-making (EAV state vector to supervised policy) under controlled oracle input from structured OULAD data; high fidelity reflects feature-oracle alignment, not general high-stakes-AI capability. The most robust finding is the intervention-bias contrast, not the absolute accuracies. We also show an Evaluation Gap: LLM-as-judge scoring (DeepEval G-Eval) is blind to intervention bias, rewarding fluent over-prescription rather than decision quality.

Craig Atkinson

Verificate Pty Ltd, Sydney, Australia

AI in Education LLM Decision-Making Language Model Bias Evaluation

Jun 27, 2026cs.CL

Can LLMs Hire Fairly? Racial Bias in Resume Screening

We audit fourteen mainstream large language models (LLMs) for hiring discrimination using the paired-resume methodology of Kline, Rose, and Walters (2022). The sole 2023-vintage model reproduces the pro-White callback gap documented in field experiments on labor market discrimination (

+2.12

pp, significant at the 1% level). Every model released in 2024 or after shows either a null gap or a significant pro-Black reversal (up to

-3.01

pp). The same pattern holds on the gender axis. Based on 24,024 paired postings per model across 14 models, our results document a reversal in the direction of algorithmic hiring bias across model generations.

Zhenyu Gao, Wenxi Jiang, Yutong Yan

Department of Finance, CUHK Business School, The Chinese University of Hong Kong

Gender Bias in Language Models Social Bias in Language Models Algorithmic Fairness Language Model Bias Evaluation AI and Labor

Jun 26, 2026cs.CL

Position Bias Correction is Insufficient for One-Pass Attention Sorting

Long-context language models suffer from position bias, where information in middle positions is underutilized. Attention Sorting addresses this by iteratively reordering documents based on attention patterns, but its multiple sort-and-generate cycles increase deployment cost. We hypothesize that position bias is the primary bottleneck and propose Debiased One-Pass Attention Sorting, which estimates a per-prompt position-bias curve from the low-attention majority of documents and uses it to correct raw attention scores (via subtraction or division) to enable single-pass sorting. Our experiments on two models refute this hypothesis in the tested setting: on LLaMA-2-7B-32K-Instruct, debiasing produces identical results to uncalibrated single-pass sorting (94.83% containment accuracy), while on YaRN-Llama-2-7b-64k, debiasing improves accuracy by 8.67 percentage points but remains 14.84pp behind iterative sorting, closing only 37% of the gap. These results suggest that position-bias correction is insufficient to match iterative sorting, and that repeated reordering provides additional benefits beyond bias correction.

Qiong Tang, Xiangkun Hu, Xiangyang Liu +2

Analemma

Long-Context Language Modeling Language Model Bias Evaluation Position Bias

Jun 23, 2026cs.CL

To Compare, or Not to Compare: On Methodological Practices in Evaluating Social Bias

As Large Language Models are increasingly deployed in critical applications, robustly evaluating their social biases is paramount. However, the current literature suffers from widespread methodological fragmentation, which yields contradictory conclusions. This stems largely from ignoring the structural framing of benchmark-level evaluations. To resolve this, we introduce a unified and controllable framework that standardizes heterogeneous benchmarks to systematically contrast isolated demographic assessments with forced-choice comparative settings. Crucially, this allows us to disentangle the confounding effects of Chain-of-Thought reasoning, neutral fallback options, and other structural artifacts in social bias evaluations. Our evaluation across multiple model families reveals a massive, systematic paradigm gap: while isolated assessments limit prejudice activation, comparative settings act as aggressive catalysts for latent discrimination, a shift primarily driven by underspecified contexts. Alarmingly, CoT reasoning exacerbates social biases under comparative settings, and this systemic bias persists as a deterministic prejudice even when models are provided neutral fallback options or claim to answer randomly. Finally, we demonstrate that this comparative prejudice is a generalized phenomenon that scales positively with model size. Ultimately, we offer a crucial methodological guideline: while researchers must leverage comparative settings to robustly audit hidden biases, practitioners cannot safely rely on comparative deployments in ambiguous real-world tasks.

Federico Marcuzzi, Xuefei Ning, Roy Schwartz +1

INSAIT, Sofia University “St. Kliment Ohridski”, Bulgaria · Tsinghua University, China · The Hebrew University of Jerusalem, Israel +1

Pairwise Comparison LLM Auditing Social Bias in Language Models Language Model Bias Evaluation

Jun 22, 2026cs.CL

UnBias-Plus: Detect, Explain, and Rewrite Bias

Bias in natural language remains a persistent challenge in both human-written and AI-generated content, affecting domains such as journalism, education, and AI research. Most existing detection methods identify only the presence of bias, with limited support for granular detection, interpretable explanations, neutral rewriting, and openly available trained models. We present UnBias-Plus, an open-source toolkit unifying (1) segment-level multi-class bias classification, (2) biased span localization, (3) neutral text rewriting, and (4) reasoning for each decision. Available via Python, CLI, REST API, and web interfaces, UnBias-Plus supports accessible bias analysis. The toolkit, source code, models, datasets, and documentation are publicly available.

Ahmed Y. Radwan, Ahmed ElKady, Sindhuja Chaduvula +3

Vector Institute for Artificial Intelligence, Toronto, Canada · Independent Researcher

Explainable Artificial Intelligence Social Bias in Language Models Language Model Bias Evaluation

Jun 22, 2026cs.LG

Memory Contagion: Cross-Temporal Propagation of Evaluator Bias via Agent Memory

Large Language Model (LLM) agents increasingly rely on memory systems to maintain long-term coherence. Recent work shows that agent memories degrade during continuous consolidation. However, existing research assumes memories are derived from unbiased experiences. In this work, we identify and formalize a novel phenomenon: Memory Contagion -- the cross-temporal propagation of evaluator bias through agent memory. We show that when agents are trained or guided by biased evaluators, their experiences become biased; when these trajectories are stored and consolidated into memory, the bias propagates to future agents retrieving from the same memory store, even when consolidation is perfect (oracle). Across two bias types (length preference, authority bias) and four experimental phases, we demonstrate: (1) Memory Contagion occurs for length bias even with perfect consolidation on older models (Gamma_A = 13.18, DeepSeek V4-Chat), while newer models (V4-Pro, Claude) are immune, proving both that biased input is a sufficient cause and that contagion is model-generation-dependent; (2) authority bias fails to propagate in all 15 controlled multi-seed experiments (Gamma_A = 0.00), revealing that not all evaluator biases can cross temporal boundaries through current memory architectures; (3) No observed safe threshold: length bias propagation is detected at contamination rates as low as p=0.2. Our findings expose a critical but contingent vulnerability in current agent memory designs and provide formal tools for measuring cross-temporal bias propagation.

Zewen Liu

Independent Researcher

LLM Agent Memory Language Model Bias Evaluation

Jun 22, 2026cs.CL

Same question, different history: language, national identity, and credit in large language models

Who invented the radio, Russia's Alexander Popov or Italy's Guglielmo Marconi? Was the telephone the achievement of Bell in the United States or Meucci in Italy? Does printing belong to China's Bi Sheng or Germany's Gutenberg? The answer depends not only on historical record but also on language and perspective. We analyse eleven widely used large language models across 21 disputed inventions and discoveries, evaluated in twelve languages and 75,896 responses. While models generally acknowledge that credit is contested, query language systematically affects which claimant is surfaced. Lower-status claimants are more likely to appear when questions are asked in their associated language, whereas dominant Anglophone figures remain stable across languages. These patterns persist after controlling for response length, model differences, historical prominence, and levels of national commemoration. Language thus acts as a switch that activates different national versions of the same history, producing systematically different national memories from the same question. We interpret this as evidence that large language models function as distributed systems of cultural memory, where language conditions which histories become visible, contributing to a computational form of banal nationalism.

William Guey, Pierrick Bougault, Wei Zhang +2

Department of Industrial Engineering, Tsinghua University, Beijing, China · School of Social Sciences, Tsinghua University, Beijing, China · Department of Industrial Engineering, Federal University of Rio de Janeiro, Brazil

Social Bias in Language Models Cultural Bias in Language Models Factual Knowledge in Language Models Language Model Bias Evaluation Linguistic Bias in Language Models Cultural Knowledge in Language Models

Jun 22, 2026cs.CL

Language-Specific Sentiment Polarity Biases in Encoder and Large Language Model Classification of Product Reviews

This study investigates sentiment polarity biases, specifically, differences in how accurately AI models classify positive versus negative reviews across languages and model architectures. Large language models show a negative bias in French and are more accurate on negative reviews, while encoder models exhibit positive bias in Japanese, missing negative reviews that use indirect criticism. These language-specific polarity biases have implications in both social and business domains deploying multilingual sentiment analysis systems.

Advita Rajiv, Kavitha Kothur, Gautham Reddy

John P. Stevens High School, Edison, NJ 08820 · Ford Foundation, New York, NY 10017 · NC State University, Raleigh, NC 27606

Multilingual Language Model Evaluation LLM Evaluation Sentiment Analysis Language Model Bias Evaluation Linguistic Bias in Language Models

Jun 19, 2026cs.CL

Per-Entity Bias Mapping for AI Visibility: Why Brand Mentions Require Entity-Specific Calibration

AI-mediated answer systems increasingly determine how brands and organizations are represented to users. Existing approaches reduce visibility to mention rate or citation frequency. This paper argues that aggregate metrics are insufficient because entities exhibit systematically different AI visibility error profiles. We introduce Per-Entity Bias Mapping (PEBM): a ten-dimensional framework distinguishing raw from verified mentions. Three failure modes are identified: (1) underrepresented entities suffer invisibility due to weak knowledge graph presence; (2) large entities suffer the Brand Hallucination Paradox -- model familiarity creates stronger surfaces for plausible but incorrect completions; (3) CEE entities face a structural infrastructure gap across knowledge graphs, NER, and entity linking. A fourth dimension, Parametric-Retrieval Lag Asymmetry, describes divergence between retrieval-augmented and parametric memory update cycles. A full-scale empirical study (n=100 Hungarian B2B entities, 1,400 probe runs, 2,062 sources) finds Tier 1 brands produce 52.69% fabricated citations versus 37.87% for Tier 3 entities (+14.82 pp; p=1.67e-11), supporting the Brand Hallucination Paradox. Regulatory-framed queries elevate fabrication to 56.77% versus 37.59% baseline (+19.2 pp). We identify rejection-induced confabulation escalation: agentic quality filters function as hallucination accelerators in compliance contexts. We introduce ghost cartography as a unifying mechanism: entities in sparse latent regions produce confident output interpolated from neighboring dense regions, yielding a two-dimensional confabulation space (fabricated presence vs. frozen representation).

Zoltan Varga

Neural Awareness Budapest, Hungary

Hallucination in Language Models Language Model Bias Evaluation

Jun 18, 2026cs.CL

Self-Preference Is Weak or Absent in Verifiable Instruction-Following Revision: A Four-Model Test Under Genuine Authorship

Large language models (LLMs) increasingly review and revise text, including their own. A documented self-preference bias (models favoring their own generations when acting as judges) raises the question of whether models also resist valid corrections to their own writing. We test this in a setting where "valid" is decided not by another model but by a deterministic verifier: instruction-following revision on IFEval. A model writes a draft; the official IFEval checker confirms the draft violates a constraint and that a candidate edit fixes it; the model then accepts or rejects that edit either as the genuine in-context author or as a fresh model that sees the draft neutrally. Across four mid-tier model families and 85 author-versus-fresh comparisons, we find no detectable self-preference: authors reject verified-good fixes to their own drafts at essentially the same rate as fresh models judging the same drafts (gap -5.1 pp, 95% CI [-12.9, +2.7]). A self-skepticism hint from a smaller pilot did not replicate at scale. The one robust observation is qualitative: when authors do reject a verified-good fix, 97% of their stated reasons are flaw-catching rather than preference, that is, about the character of rejections, not an elevated rate. Effects smaller than ~13 pp cannot be excluded at this sample size.

William Guey, Pierrick Bougault

Department of Industrial Engineering, Tsinghua University, Beijing, China

LLM Self-Correction Language Model Bias Evaluation

Jun 17, 2026cs.CL

Reliability without Validity: A Systematic, Large-Scale Evaluation of LLM-as-a-Judge Models Across Agreement, Consistency, and Bias

LLM-as-a-Judge has become the dominant evaluation paradigm for language models, but judge validation in practice relies on exact-match agreement, a metric that does not correct for chance and systematically overstates discriminative ability. We present the largest systematic evaluation of LLM-as-a-Judge to date: 21 judges from nine providers across MT-Bench, JudgeBench, and RewardBench, evaluated under three protocols (agreement, consistency, bias audit) over 118 runs and approximately 541,000 individual judgments. Four findings emerge, consistent across the full cohort, including the April 2026 frontier: kappa deflation between exact match and Cohen's kappa is universal (33--41 pp on MT-Bench), judge rankings shift by up to 14 positions across benchmarks, high test--retest reliability (>0.95) coexists with severe position bias (>0.10) in two production-deployed judges (instantiating a consistency--bias paradox), and verbosity bias is small (<0.011) across our cohort under a single pairwise rubric. We distill these into a Minimum Viable Validation Protocol.

Justin D. Norman, Michael U. Rivera, D. Alex Hughes

UC Berkeley School of Information 102 South Hall Drive, Berkeley CA, 94708

Inter-Rater Reliability LLM Evaluation LLM-as-a-Judge LLM Reliability Language Model Bias Evaluation Benchmark Validity+1

Jun 17, 2026cs.LG

MortarBench: Evaluating Mortgage Loan Origination Agents

Loan origination is the process by which a lender creates a new loan, from application and underwriting through approval and funding. This process serves a critical role in evaluating the eligibility and level of risk posed by an applicant. Recently, firms have begun using mortgage loan agents to augment human loan officers, despite a lack of any public benchmark. To fill this gap, we present MortarBench, a loan origination agent benchmark. MortarBench uses a financial data synthesis and mutation pipeline to generate examples with broad edge case coverage that match real-world distributions and questions. We find that state-of-the-art large language models (LLMs) perform poorly, with closed-source models achieving at most 77.1% exact match accuracy. We also discover systematic biases in LLM perception of foreignness related to non-English names. Noting these weaknesses, we introduce CRIT, a confidence calibration framework. Our method increases accuracy to 80.5% while improving risk management steering and reducing bias.

Matthew Toles, Yunan Lu, Manav Munjal +6

Columbia University · Tidalwave

Financial Services Language Model Calibration Language Model Bias Evaluation Finance Agent Benchmarks

Jun 17, 2026stat.ML

Quantifying and Auditing LLM Evaluation via Positive--Unlabeled Learning

Large Language Models (LLMs) are increasingly used as judges for scalable evaluation, yet such LLM--as--a--Judge systems exhibit systematic biases that are decoupled from semantic quality, most notably verbosity bias. Meanwhile, human supervision is costly and typically selective, yielding reliable positive judgments but leaving most outputs unlabelled and potentially mixed in quality. We formulate LLM evaluation under selective human supervision as a positive--unlabelled learning problem and propose a geometric auditing framework based on Partial Optimal Transport. By aligning a small set of human--verified positives with a reliable subset of unlabelled outputs in a fixed embedding space, our method identifies human--consistent preferences and corrects biased judges without retraining. Experiments demonstrate improved alignment with human preferences, increased robustness to presentation biases, and interpretable confidence estimates, offering a scalable and statistically grounded alternative to existing LLM--as--a--judge pipelines.

Zilong Zhang, Yi-Ting Hung, Lei Ding +1

Department of Mathematics and Statistics Georgia State University · Department of Statistics University of Manitoba

Inverse Optimal Transport LLM-as-a-Judge LLM Auditing Language Model Bias Evaluation Bias in LLM-as-a-Judge Positive-Unlabeled Learning

Jun 17, 2026cs.CL

The Wrong Kind of Right: Quantifying and Localizing Misfired Alignment in LLMs

Warning: This paper studies stereotypes and biases, and contains potentially disturbing examples, used for illustration purposes only. Our findings should not be interpreted as an argument against alignment. Instead, this paper highlights the need for principled approaches to more advanced alignment. Alignment aims to ensure that large language models (LLMs) behave safely and reliably, including by avoiding unsafe inferences. However, we show that such safety-oriented behaviors can misfire: models may reject warranted conclusions even when they are explicitly supported by context. We call this failure mode misfired alignment, where alignment-induced changes cause LLMs to override explicit evidence. To quantify this phenomenon, specifically on stereotype-related alignment, we introduce VETO, a benchmark consisting of 2,032 BBQ-derived contrastive pairs, and define a new metric, Misfired Alignment Rate (MAR), which measures on a 0 to 100 scale how often a model fails on a stereotype-related question but succeeds on its contrastive counterpart. We benchmark 25 LLMs on VETO, and show that all LLMs, including the most recent ones, exhibit non-trivial (4.7 to 18.9%) MARs while all human participants achieve 0.0% MAR. Controlled priming experiments further show that alignment-induced cues can substantially amplify MAR across LLMs, indicating that these failures are not merely artifacts of individual examples but can be induced by safety-related framing. Mechanistic analyses on open-weight LLMs reveal late-layer suppression of evidence-supported answers, and comparisons between instruct and base LLMs suggest that this suppression emerges after instruction training. These findings show that current alignment methods can overgeneralize surface-level safety cues, to the point of overriding objective evidence, motivating more work on alignment objectives that better preserve contextual grounding.

Naihao Deng, Yiming Feng, Chimaobi Okite +4

University of Michigan · University of Cambridge · University of Aberdeen

LLM Grounding LLM Alignment Language Model Safety Evaluation Social Bias in Language Models Language Model Bias Evaluation

Jun 17, 2026cs.MA

Gender Bias in LLM Hiring Decisions: Evidence from a Japanese Context and Evaluation of Mitigation Strategies

Large language models (LLMs) are increasingly deployed in hiring workflows, yet most research on gender bias in LLM hiring decisions has focused on English-language, Western-format resumes. This study examines whether pro-female gender bias extends to a Japanese corporate context and evaluates two practical mitigation strategies. Using a counterfactual resume design with 60 Japanese rirekisho-format resumes, 12 name pairs selected on linguistically grounded gender-signal criteria, and five state-of-the-art LLMs (Claude Sonnet 4.6, GPT-4o, DeepSeek-V3, Gemini 2.5 Flash, Llama 3.3 70B), we conducted 43,200 API calls across baseline, prompt instruction, and privacy filter conditions. A crossed random-effects linear mixed model confirms a significant pro-female bias across all five models, replicating Western findings in a non-Western context. A prompt-level gender-neutrality instruction produces no meaningful reduction in bias. A name-reliance analysis formally identifies the candidate name as the primary gender channel: removing the name from the prompt reduces the female effect by nearly its full magnitude. An unexpected incompatibility between the privacy filter and GPT-4o's content safety filter, resulting in a 42% refusal rate, highlights a practical deployment challenge for name anonymization in LLM-assisted recruitment pipelines.

Serena A. Hoffstedde, Machiko Hirota, Akshara Nadayanur Sathis Kanna +4

Shibaura Institute of Technology, Tokyo, Japan · Amsterdam University of Applied Sciences, Amsterdam, Netherlands · University of Pennsylvania, Philadelphia, USA +2

Gender Bias in Language Models AI-Assisted Decision Making Language Model Bias Evaluation

Jun 16, 2026cs.CL

Unintended Effects of Geographic Conditioning in Large Language Models

Modern conversational AI systems frequently rely on user metadata to localize responses, yet the unintended regional biases introduced by this hidden context remain poorly understood. In this work, we evaluate location leakage: the phenomenon where a model generates geographic references despite receiving a geographically neutral user prompt. Across both creative writing and open-ended Q&A prompts, even state-of-the-art LLMs systematically favor region-specific outputs when exposed to location metadata, with leakage spiking by up to 793 times above baseline (e.g., from 0.04% to 31.7% for Llama 3.1-8B, and 21.3% and 8.8% for Qwen3-8B and Claude Sonnet 4.6, respectively). Our analysis further shows a novel structural conditioning effect: replacing the injected location with the placeholder "Unknown" still elevates leakage by up to 72 times above baseline, demonstrating that the user profile frame itself, independent of any geographic content, acts as a generative conditioning signal.

Naz Col, David M. Chan

University of California, Berkeley

Privacy Leakage in Language Models Language Model Bias Evaluation

Jun 16, 2026cs.CL

LLMs Infer Cultural Context but Fail to Apply It When Responding

Recent work has shown that LLMs overrepresent dominant cultures, particularly Western ones, while marginalizing others. We investigate whether this affects models' ability to generate culturally adapted responses by evaluating their use of local measurement units based on the user's perceived cultural background. We introduce Cultural and Pragmatic Response Inference (CAPRI), a dataset of conversations with varying levels of cultural cues. Experiments with state-of-the-art LLMs show that models can infer cultural background and recall relevant conventions, but often fail to utilize the information to adapt their answers to the relevant cultural conventions, unless explicitly prompted to perform the tasks sequentially. We further evaluate adaptation to the interpretation of time and quantity expressions, two subjective language grounding dimensions that are affected by culture. We find that models increasingly adapt their answers as cultural cues accumulate, but their priors are not culture-neutral, sometimes aligning with the model's country of origin. Overall, CAPRI provides a resource for future research aimed at narrowing the gap between cultural knowledge and culturally adaptive language generation.

Yisong Miao, Jian Zhu, Vered Shwartz

National University of Singapore · University of British Columbia · Canada CIFAR AI Chair, Vector Institute

Pragmatic Reasoning in Language Models Cultural Bias in Language Models Language Model Bias Evaluation Cultural Knowledge in Language Models

Jun 16, 2026cs.CL

Evaluating Second-Order Bias of LLMs Through Epistemic Entitlement

Evaluations of social bias in LLMs largely focus on whether models generate or imply biased content. However, as LLMs are increasingly used as judges of bias, they may exhibit social biases in subtler ways in how they evaluate biased content, which current methods do not systematically capture. We call this second-order bias: social bias in an LLM's judgment about social bias, which we evaluate through a novel, philosophically grounded reasoning task. Drawing on entitlement epistemology, we conceptualize bias as misplaced foundational knowledge that shapes an agent's rational inquiry, and derive a logical reasoning task for LLMs to judge to whom a biased text is acceptable or non-acceptable. We develop two simple metrics to measure how biased LLM judges are in inferring demographics for acceptability without sufficient support, and how these inferences vary across groups targeted by biased texts. Evaluating open and closed models, we find that our task evades safety guardrails by surfacing bias in model judgment. It varies systematically across target groups, reflects implicit social maps, and shows how models are still triggered by demographic labels. Our work points to the need for LLM bias evaluation in judgment tasks and broadly, for more theoretically grounded approaches to bias evaluation in NLP. We release our code and model responses at https://github.com/uofthcdslab/second-order-bias.

Ramaravind Kommiya Mothilal, Terry Jingchen Zhang, Raiyan Ahmed +3

University of Toronto · Vector Institute · EuroSafeAI +1

LLM-as-a-Judge Social Bias in Language Models Language Model Bias Evaluation Bias in LLM-as-a-Judge

Jun 16, 2026cs.AI

Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems

Large language models (LLMs) are becoming a major way for consumers to find products, but we do not yet understand how brands compete in this new channel. We study brand dynamics in LLM recommendations using skincare products -- a category where consumers cannot easily judge quality before buying and must rely on brand reputation -- across three commercial LLMs (GPT-4o-mini, Claude Sonnet, Gemini 3 Flash), with a robustness check on search goods. In three experiments, we find: (1) a Conditional Monopoly where well-known brands get recommended 100% of the time (IAI = 10.0) when all products have the same specifications, but this dominance disappears with less than a +0.1-star rating advantage for a competitor; (2) authority-style marketing language, including fabricated clinical-evidence claims, breaks this monopoly at a Bias Surplus Value equal to +0.17 rating points, with each model responding differently; and (3) a social dilemma in multi-brand GEO competition: when all brands adopt the same optimization strategy, individual payoff falls from +0.802 to +0.007 in our payoff proxy, and non-participating brands receive zero recommendations in our tests. Our results suggest that generative engine optimization (GEO) should be studied not only as a security risk, but also as an emerging marketing practice that shapes market competition.

Xi Chu, Yupeng Hou

Trine University · Texas A&M University

Generative Engine Optimization Language Model Bias Evaluation Large Language Model-Based Recommendation

Jun 15, 2026cs.CL

Thinking Like a Scientist? A Structural Study of LLM-Generated Research Methods

Large Language Models (LLMs) are increasingly used to guide research methodology, yet their default methodological tendencies under minimal prompting remain unclear. Here, we prompt GPT-5.1, Gemini 3 Pro, and DeepSeek-V3.2 with an LLM-extracted research question from each of 1,000 recent arXiv computer-science papers and compare the resulting methodology suggestions against a paper-derived experimental inventory. Since we provide only the research question, the differences we measure reflect initial suggestions and not how optimal those suggestions are. We extract structured method features from both sources, map them into a shared taxonomy, and quantify divergence across multiple taxonomy dimensions including model provider, dataset task type, and evaluation metric type. The strongest imbalance appears in provider choice, with Jensen-Shannon divergence about 3-5x larger than any other taxonomy dimension. Other/Academic single-occurrence models are underrepresented by 23-24 percentage points, while reused academic/community models are slightly overrepresented (4-6pp). LLMs also suggest a much narrower range of methods overall: the effective number of model entities contracts from 1,232 to 59-96, and inter-LLM rank correlations (0.55-0.68) generally exceed LLM-to-paper correlations (0.33-0.56), so the distortions are largely shared across models. Popularity baselines, BM25 retrieval calibration, and paper-level similarity tests confirm that the outputs are query-specific responses, but filtered through a narrower set of options. Researchers who rely on LLM suggestions without cross-checking therefore risk narrowing their methodological search space toward a more concentrated default.

Francesca Carlon, Brecht Verbeken, Vincent Ginis +1

1Data Analytics Lab, Vrije Universiteit Brussel, Pleinlaan 5, 1050 Brussels, Belgium · 2imec-SMIT, Vrije Universiteit Brussel, Pleinlaan 9, 1050 Brussels, Belgium · School of Engineering and Applied Sciences, Harvard University, Cambridge, Massachusetts 02138, USA

LLM Evaluation AI-Assisted Scientific Research Language Model Bias Evaluation

Jun 15, 2026cs.CL

P3B3: A Multi-Turn Conversational Benchmark for Measuring European and Brazilian Portuguese Variety Bias in LLMs

As Large Language Models (LLMs) become embedded in everyday communication, capturing regional linguistic variation is essential for reliable and equitable language use. In Portuguese, European (pt-PT) and Brazilian (pt-BR) varieties remain unevenly represented, with pt-BR dominating in data quantity, while LLM preference for Portuguese variants remains underexplored. To address this gap, we introduce P3B3, an expert-curated language variety agnostic benchmark of conversational prompts, along with an evaluation framework for measuring variety bias and controllability. Experiments on several models show that most LLMs exhibit a strong bias toward pt-BR, with variation in controllability across models. These results highlight the need for more balanced multilingual representation across language varieties.

Rafael Ferreira, Inês Vieira, Inês Calvo +6

NOVA University of Lisbon, Portugal · NOVA LINCS

Multilingual Language Model Evaluation Social Bias in Language Models Language Model Bias Evaluation Linguistic Bias in Language Models

Jun 15, 2026cs.LG

MIRAGE: Auditing Anti-Muslim Bias in Frontier LLMs Across Reasoning, Agentic, and Time-Coupled Conditions

Five years after the discovery of persistent anti-Muslim bias in large language models, most evaluations remain confined to single-turn prompt completion, a setting that no longer reflects how frontier LLMs are deployed. We introduce \textbf{MIRAGE} (Muslim-Identity Reasoning and Agentic Generation Evaluation), a benchmark of 1{,}200 prompts spanning three deployment-realistic conditions: direct completion, chain-of-thought reasoning, and simulated agentic decision-making across content moderation, lending triage, refugee claim summarization, and hiring screens. Across six frontier models, we find that (i) chain-of-thought reasoning \emph{amplifies} rather than suppresses Muslim-violence associations by 12--34% relative to direct completion, (ii) agentic decisions exhibit a 9--22 percentage-point asymmetry between Muslim and matched non-Muslim cases on identical evidence, and (iii) bias is sharply time-coupled to retrieved news context, increasing 18--27% under recent-conflict retrieval. Existing prompt-based mitigations transfer poorly across our three conditions, suppressing direct-completion bias while leaving agentic asymmetry largely intact. We release MIRAGE and an open evaluation harness to support targeted mitigation research.

Noor Islam S. Mohammad, Tamim Sheikh

Department of Computer Science, Informatics Institute, Istanbul Technical University, ˙Istanbul, Türkiye · Department of Computer Science and Engineering, Jashore University of Science and Technology, Bangladesh.

LLM Safety Benchmarks LLM Auditing Social Bias in Language Models Language Model Bias Evaluation Linguistic Bias in Language Models