Position Bias

Follow this topic Ask about this topic

Momentum

5 papers in the last four weeks, against 2 the four weeks before. 0.1% of all new papers.

Jul 6Week of Sep 21

Latest papers 53

Sep 30, 2026cs.IR

A Shared Taste for Model-Written Text: The Generator-by-Selector Matrices of "AI-AI Bias" Show No Detectable Own-Model Premium

Laurito et al. (PNAS 2025) showed that large language models choosing between two descriptions of the same product, paper or film prefer the description written by a language model over the one written by a person, by a wide margin over what human judges do. Their design crosses five generators with the same five models as selectors, which permits a second question the paper does not headline: does a selector prefer text from its own model beyond what the generator and selector main effects predict? We rebuild the three 5x5 matrices from the per-item counts in the authors' public repository (21,828 valid trials; every cell matches the published value) and fit a two-way fixed-effects model with an own-model term gamma, tested by the exact permutation test over the 120 relabellings of the selectors. The premium is +0.013 on products (exact one-sided p = 0.24), -0.010 on paper abstracts (p = 0.74), +0.054 on films (p = 0.07) and +0.019 pooled (p = 0.14; 95% interval -0.008 to 0.046). The same-vendor term for the GPT-3.5 and GPT-4 pair is negative in all three datasets. Position bias moves single cells by up to 0.42 share points in either direction, and the own-model contrast is unchanged once order-driven items are removed. The design would have detected a premium of 0.05 with 82% (products), 88% (papers), 42% (films) and 97% (pooled) power; the minimum detectable effect at 80% power is 0.034 pooled. The absence is informative down to about 0.04 share points and silent below that. The 4x4 matrix of Tan et al. (ACL 2024) gives gamma = +0.148 at the smallest p its 24 relabellings allow, with a same-family term of the same size. The main result of Laurito et al. stands: models share a taste for model-written text, with GPT-4's descriptions chosen 77% to 95% of the time by every selector on products. What these data do not show is a model recognising and favouring its own prose.

Dmitrij Żatuchin

Department of Information Technologies, Estonian Entrepreneurship University of Applied Sciences (EUAS), Tallinn, Estonia. · Rankfor.AI, Wroc law, Poland; Tallinn, Estonia.

Large Language Model Bias Human-Written Text Position Bias Contrastive Learning Matrix

Sep 30, 2026cs.AI

More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models

Direct-decision models turn text into low-latency structured labels and scores, making them attractive for classification and automatic evaluation. Yet reliability requires more than accuracy: a model must also use the ordinal decision scale supplied by the user faithfully. We analyze JEV~1.13 and three open KEV models. Our investigation begins with ANLI, where JEV assigns 38.8% of all predictions and 51.3% of errors to Neutral despite 74.95% accuracy, nearly balanced gold labels, and balanced candidate positions. Across 36 ordinal datasets, final decisions use only 67--76% of the effective gold support, versus 87--102% on four nominal tasks. Randomizing candidate order weakens but does not remove this compression. Holding items and source scores fixed while balancing gold support and positions, we refine scales from

K=2

to

14

; utilization falls for every model and reaches 26--75% at

K=14

, although candidate probabilities remain broad for most models. Targeted BA-LoRA post-training raises gold-relative utilization from roughly 47% to 86% on eight supervised scales at both KEV sizes, showing that the compression is learned and modifiable rather than an immutable architectural limit. We call this ordinal scale-utilization bias: decision-stage candidate-space compression distinct from accuracy, gold imbalance, fixed position, and candidate count alone. The code and data are available at https://github.com/Glax147/jev_ordinal_scale_bia

Tianxiang Gao, Jinzhe Li, Zhiyuan Li +2

School of Artificial Intelligence, Jilin University · College of Software, Jilin University · International Center of Future Science, Jilin University +1

Position Bias Order Matters Decisions

Sep 29, 2026cs.CL

Pair Difficulty Matters: Rethinking Pairwise LLM-as-a-Judge Evaluation and Consistency

Large Language Model judges are widely used to rank texts and text-generating systems through pairwise comparison, and their reliability is typically assessed via three proxies: position bias, transitivity, and pairwise agreement (self- or human-labeled). Because these proxies drive judge selection and benchmarking, a substantial literature reporting that judges perform poorly on them risks steering practitioners away from otherwise capable evaluators. We argue this assessment is misleading. Under the Bradley--Terry geometry underlying pairwise aggregation, each proxy is dominated by close-rank-gap pairs, where inconsistency is information-theoretically expected and individual verdicts contribute little to the aggregate ranking; far-gap pairs carry the ranking signal but barely move the proxies. We formalize this argument and validate it in a controlled simulation and on two human-rated corpora: the proxies correlate only weakly with ranking accuracy against gold, and their predictive component concentrates in the far-gap regime. Judges should therefore be assessed on rank-gap-conditional metrics, ideally against human rankings. Code at https://github.com/brunobrocai/PairDifficulty.

Bruno Brocai, Maria Becker

Heidelberg University German Department

Large Language Model Judges Llm-As-A-Judge Ranking Position Bias Text Corpora

Sep 29, 2026cs.LG

Unlocking the Critic: Reward-Free Policy Optimization for LLM Post-Training

Recent approaches to reinforcement learning (RL) post-training for large language models increasingly remove the critic to reduce training instability and memory overhead. Even where a critic is trained, it is discarded once training ends, although it has learned to predict outcomes. We revisit this trend and show that a pretrained critic's ability to predict future outcomes can make it a valuable asset for efficient long-horizon reasoning. First, we find that instability in critic-based RL for long chain-of-thought reasoning is largely an optimization artifact: keeping policy updates small and low in variance restores stable convergence. Second, a well-pretrained critic estimates the posterior probability of eventual success from later trajectory states and unfinished prefixes. Its predictions provide outcome-derived, dense, per-prefix learning signals that, during policy optimization, require neither completed rollouts, step-level annotations, nor external reward labels. Building on this insight, we introduce Reward-Free Policy Optimization (RFPO), which repurposes a single calibrated, frozen critic as a rollout-level reward, a value baseline for generalized advantage estimation, and a success forecaster for unfinished prefixes. We further show that binarizing the debiased score stops the policy from exploiting the critic's length bias. Binarized, RFPO matches supervised PPO without a single label in the training loop, while cutting compute and memory overhead. This makes RFPO well suited to long-horizon reasoning tasks, where outcomes arrive late and generation dominates cost: because rollouts can be rewarded before they finish, training no longer has to pay for waiting on every trajectory to complete. Our findings challenge the prevailing critic-free paradigm and establish critic-based, reward-free optimization as a scalable and computationally efficient path for LLM post-training.

Hongyang Li, Xiao Li, Caesar Wu +3

University of Luxembourg · Seafill Open-Source Community · Université Paris-Saclay

Critic-Free Reinforcement Learning Reward-Only Policy Optimization Variants Critic Long-Horizon Task Planning Position Bias

Sep 28, 2026cs.IR

Recommendation Ranking Off-Policy Evaluation under Ranking-Dependent Examination via Examination-Relevance Decomposition

Off-policy evaluation, which estimates evaluation policy performance from logged data, is key for recommender ranking policies. However, logged clicks cannot distinguish unexamined items from examined non-clicks, causing bias in existing estimators when the assumed examination structures fail. We propose two estimators based on the decomposition of clicks into examination and relevance. First, the latent-examination independent inverse propensity score (LE-IIPS) estimator corrects the IIPS bias using policy examination probability ratios. Second, the examination-decomposed doubly robust (ED-DR) estimator extends LE-IIPS to a doubly robust framework. ED-DR is unbiased if the examination probabilities are correct regardless of relevance accuracy, or under ranking-independent examination, even if both model estimates are inaccurate. Experiments show that ED-DR achieves a lower MSE than existing methods with large sample sizes, especially when the examination depends on ranking. We also highlight its limitations under small samples or cascade user behavior conditions.

Riki Okamura, Toshiharu Sugawara

Department of Computer Science and Communications Engineering, Waseda University, Tokyo, Japan

Policy Evaluation Ranking Multi-Dimensional Evaluation Off-Policy Learning Assessment Probability+1

Sep 15, 2026cond-mat.dis-nn

Bias-Induced Crossover in Absolute Capacity of Dense Associative Memory

The absolute capacity of dense associative memory has mainly been analyzed for unbiased patterns. Here we examine the effect of bias in centered binary patterns under the Krotov-Hopfield single-site criterion

P_{\mathrm{error}}=1/N

, where

P_{\mathrm{error}}

is the probability that a single-site flip lowers the energy of a stored pattern and

N

is the number of neurons. Each pattern component takes

1-q

with probability

q

and

-q

otherwise, where

0<q\le1/2

. For polynomial interactions of order

n

, a signal-to-noise analysis gives an absolute capacity of order

N^{n-1}/\ln N

at

q=1/2

. For fixed

q<1/2

, however, the capacity is

O(N^{n/2})

for even

n\ge4

and

O(N^{(n+1)/2})

for odd

n\ge5

. For

n=3

, both the unbiased and fixed-bias capacities remain

O(N^2/\ln N)

. For

n\ge4

, these different asymptotic forms imply a nonuniform large-

N

limit near

q=1/2

. Asymptotic matching predicts a bias-induced crossover in the region

1-2q=O(\ln N/N^{\lfloor n/2\rfloor-1})

. The crossover originates from a bias-dependent crosstalk mean that reduces the stability of sites carrying the more frequent value

-q

. Computer simulations are compared with the finite-size conditioned-Gaussian predictions. An activity-dependent control potential that cancels the conditional crosstalk mean restores the

N^{n-1}/\ln N

capacity for fixed

0<q<1/2

within the conditioned-Gaussian approximation.

Yuto Sakurai, Takeaki Shimokawa, Kazunori Iwata +1

Graduate School of Information Sciences, Hiroshima City University, 3-4-1 Ohtsuka-higashi, Asaminami-ku, Hiroshima 731-3194, Japan · RIKEN Center for Advanced Intelligence Project and The University of Tokyo

Dense Associative Memory Hopfield Networks Capacity Position Bias Bits Crossing+1

Sep 12, 2026cs.AI

A Unified Per-Token Gating Family for On-Policy Distillation: FKL/RKL Mixing with Multi-Channel and Bias Coefficients

Per-token gating of forward/reverse KL losses has become a standard technique for on-policy knowledge distillation (OPD), but existing methods such as EOPD (Jin et al., 2026) and ToDi (Jung et al., 2025) each fix a single gating signal and a single gating direction, and the two have never been compared directly. We introduce a four-coefficient parameterization lambda_t = sigma(a * h_t + b * u(x) + c + d * gap_t) in which direction-aligned proxies of EOPD and ToDi appear as one-dimensional (1D) restrictions, and which adds multi-channel composition and an explicit bias as further degrees of freedom. On TweetEval (Barbieri et al., 2020) emotion and hate, with a Qwen3-32B teacher and a Qwen3-4B student, configurations in the full family reach higher accuracy than the matched-magnitude single-channel (entropy-only / gap-only) 1D restrictions in 33 of 36 comparable cells, and a 26-cell mean-match isolation experiment places dynamic gating ahead of effective-KL-matched static baselines in 19 of 26 cells. Because cells share training data, models, and parameter substructure, we report both counts as exploratory aggregate directional evidence rather than as independent hypothesis tests. Targeted three-seed paired replications of the nine headline comparisons singled out by that sweep -- including a third task, offensive -- are directionally consistent, but individually smaller than the single-seed estimates and not significant at n=3. We therefore present the parameterization primarily as a shared coordinate system for comparing per-token gating designs in short-output classification OPD.

Suwan Wu, Yumeng Lin, Pengcheng Yuan +1

Xiaohongshu Inc. · Tianjin University

Token-Level Uncertainty Position Bias Channel

Sep 10, 2026cs.NE

Breaking the Central Bias: Spatially Partitioned Experts for Coordinate-Based Neuroevolution

Evolvable-Substrate HyperNEAT (ES-HyperNEAT), a bio-inspired indirect encoding that determines neuron placement and connection weights from spatial coordinates, exhibits a failure mode on MNIST as a diagnostic benchmark. Because input pixels map to a coordinate space centered at the origin, evolved networks converge on a small central cluster of input pixels, a spatial-concentration bias; prior work observed only 21% mean accuracy in this regime. Is this bias an optimization artifact or an architectural ceiling? Inspired by Mixture-of-Experts (MoE) principles, we partition the input into non-overlapping spatial segments, each assigned to a separately evolved specialist network. With 13 such experts, this design reaches 43% mean accuracy, a 106% relative improvement over the baseline. The architectural gain does not depend on data-driven aggregation: equal-weighted averaging, which uses no validation data, already yields a 70% improvement; the gain comes from partitioning, not the weighting. Receptive-field analysis shows the mechanism: partitioning forces evolution to discover features across the entire image, expanding active pixel coverage from 4% to 79%. Absolute accuracy stays below gradient-trained baselines, but the relative gain points to central bias, not the evolutionary search. Two tools are designed to generalize beyond MNIST: a receptive-field diagnostic for silent input-coverage collapse, and a spatial-partitioning remedy that restores coverage.

Romain Claret, Arthur Gygax, Michael O'Neill +3

University of Neuchâtel, Neuchâtel, Switzerland · University College Dublin, Ireland

Neuroevolution Mnist Receptive Fields Evolutionary Search Bio-Inspired Framework Experts+1

Sep 3, 2026cs.HC

GazeFS: Target-Centered Gaze-Trajectory Forecasting and Stabilization from Gaze-Head History

Target-centered gaze interaction requires more than suppressing frame-to-frame fluctuations: target acquisition produces task-aligned changes in gaze-head dynamics, while a gaze trace may retain a persistent target-relative residual direction. We formulate gaze correction as online target-centered gaze-trajectory forecasting and stabilization and introduce GazeFS, which maps a variable-length gaze-head history to the next target-center direction and a short-horizon Search/Focus estimate without target information at inference. Across 7,960 acquisition episodes from 30 participants, Search-Focus differences remain stable under quality control, onset exclusion, and duration matching. History windows improve phase decoding over the current endpoint, but explicit task progress remains a strong control. Under the 30-participant, five-fold grouped out-of-fold protocol across three seeds, the reductions relative to raw hold in Focus episode bias, within-episode dispersion, and P90 target error are 0.182 degrees, 0.257 degrees, and 0.400 degrees, with participant-bootstrap 95% confidence intervals excluding zero. Endpoint-free replay from empty history preserves the Focus advantage and yields raw-network phase balanced accuracy/AUPRC of 0.925/0.993; coordinate controls further show that recent history contributes beyond explicit progress metadata. GazeFS therefore improves Focus target centering and empirical residual contraction while leaving temporal smoothness as a separate objective.

Yaozheng Xia, Zaiping Zhu, Bo Pang +4

School of Information Science and Technology, School of Artificial Intelligence, Beijing Forestry University · School of Digital Art, Nanjing University of the Arts · School of Computer Science, Peking University +1

Gaze Target Estimation Stabilization Fluctuation-Dissipation Theorem Position Bias

Sep 1, 2026stat.ML

Pooling and Drift in Delayed Bandits

A system often has to act long before it learns whether the act worked: a recommender sees a click in seconds and a purchase in days. With

K

actions and a delay of

d

rounds, the best rate known for this setting is

\widetilde{O}(\sqrt{(K+d)T})

over

T

rounds, so a longer menu is always more expensive to learn from. It need not be: if the outcome depends on the action only through the state it produced, then one late outcome informs every action that could have produced the observed state, and the price is set by how many genuinely different states the actions produce rather than by how many actions there are. We measure this using an effective dimension

v_t

between

1

and the number of states, and prove

\widetilde{O}(\sqrt{(d+1)V\log K})

for a rotating algorithm and

\widetilde{O}(\sqrt{V^{-}}+\sqrt{dT})

for the single-copy algorithm used in practice, for any budget fixed in advance; merging similar states lowers the price further, at an explicit bias. Even when given the exact losses from

d

rounds ago, no algorithm escapes

Ω(\sqrt{dE\min\{1+\log J,T/d\}})

, where

J

counts the drifting directions and

E

bounds how far losses move while the learner waits. On generated data, the state channel cuts regret by up to 79 percent against action-level weighting and, on the funnel family, by 32 to 68 percent against a tuned minimax-optimal method.

Melika Baghi

Georgia Institute of Technology

Stochastic Multi-Armed Bandits Minimax Optimal Sample Complexity Position Bias Global Average Pooling

Aug 24, 2026cs.AI

Does Rank Still Matter? Position Bias When AI Agents Shop on Our Behalf

Search rankings are valuable because human attention is scarce and sequential. Higher-placed alternatives are easier to find, so they are examined and bought more often. Consumers are now delegating search to AI agents that can ingest an entire results page at once. Randomizing the order of one hundred hotel listings across 5,000 AI agent sessions, we compare four large language models against human field data. AI agents search more deeply than humans and never decline to buy. Position still predicts which listings are inspected, but weakly and non-monotonically: the middle of a results page has the lowest probability of inspection, not the bottom. Position reaches the choice stage for some models and not others, a heterogeneity that tracks neither provider nor capability. All models nonetheless converge on the same undominated listing. For agentic search, the attributes displayed on a results page matter more than placement within it.

Davood Wadi, Yu Ma

Agentic Commerce Agentic Search Ranking Artificial Intelligence Agents Position Bias

Aug 9, 2026cs.CL

Are LLMs Positionally Consistent Ordinal Classifiers? A Systematic Evaluation

Large language models are increasingly used for ordinal classification, yet semantically equivalent changes to prompt organization can alter their predictions. We conduct systematic experiments to characterize positional bias from label order, demonstration order, and demonstration placement. First, we apply the three probes to ten frontier LLMs on a common ordinal-classification task; every model is sensitive to all three positional sources, showing that the problem is pervasive. Second, we vary eight prompt-, task-, and model-level factors across five datasets; accuracy and stability are often misaligned, and only lower scale cardinality consistently improves both. Third, we compare pointwise, pairwise, and listwise inference, alternative aggregation and debiasing methods, and joint configurations; the tested corrections do not provide a reliable remedy, while a comparison-based listwise formulation offers the best balance but transfers unevenly across models and bias sources. These findings show that positional robustness depends on the full system configuration rather than the model alone. Ordinal-classification systems should therefore be selected jointly for predictive performance and stability.

Yu Wang, Zhe Zhou, Menglin Liu +1

Cornell University · University of Washington · The Chinese University of Hong Kong, Shenzhen +1

Position Bias Order Matters Position

Jul 25, 2026cs.LG

On the Impossibility of Unbiased and Length-Invariant Policy Optimization with Outcome Rewards

Group Relative Policy Optimization (GRPO) is the dominant reinforcement learning algorithm for training reasoning capabilities in large language models, notably adopted by DeepSeek-R1. The recent improvement Dr. GRPO (COLM 2025) identifies the response-level length bias caused by per-trajectory length normalization in GRPO and proposes removing this normalization, claiming the resulting optimizer is "unbiased." We show that this claim is incomplete. Specifically, we establish an impossibility theorem: under the standard outcome reward + GRPO setting, no length-based weighting scheme can simultaneously achieve the following two properties. (P1) Gradient unbiasedness: the gradient estimator is an unbiased estimate of the true policy gradient. (P2) Length invariance: each trajectory's effective contribution to the gradient is independent of its token length. GRPO approximately satisfies P2 but violates P1; Dr. GRPO satisfies P1 but violates P2. We characterize the complete tradeoff spectrum via the parametric family f_alpha(L) = L^{alpha - 1}, where alpha = 0 recovers GRPO, alpha = 1 recovers Dr. GRPO, and provide quantitative analysis showing that Dr. GRPO's length bias can cause longer trajectories to dominate gradient updates by a factor proportional to the length ratio. Our results reveal that neither algorithm is universally "done right"; they occupy opposite ends of a fundamental and unavoidable tradeoff.

Fei Ding, Yongkang Zhang, Yuhao Liao +2

Alibaba Group · Tsinghua University

Policy Gradient Frictive Policy Optimization Reward Signal Position Bias Impossibility

Jul 23, 2026cs.LG

Position Bias is Hidden Behind Ceiling Effects: A Permutation Diagnostic for LLM Benchmarks

Position bias in multiple-choice LLM evaluation is widely cited as a confound in capability comparisons, but published measurements rely on single answer-order shuffles whose results confound the bias signal with content-level noise and sampling stochasticity. I introduce inspect_permute, an open-source extension to the inspect_ai evaluation framework that runs exhaustive answer-order permutations per question and reports the chi-squared / Cramer V signature of position bias with bootstrap confidence intervals. I apply the tool across four vendors (gpt-4o-mini, claude-haiku-4-5, gemini-2.5-flash, grok-3) on five MMLU subjects, 24,000 API calls under temperature-0 generation, with falsifier predictions pre-registered via a public SHA-256 hash before half the data was observed. Position bias turns out to be statistically detectable only within a roughly 60-95% base-accuracy Goldilocks zone. Below it, processing-load dominance swamps subject-specific signal; above it, ceiling effects compress the variance below the chi-squared test resolution. Detectable cells separate into two mechanism types: monotone A-to-D decrease (processing_load, in low-tier models) and non-monotone D-drop (content_ambiguity, in a narrow capability band). Standard MMLU places every frontier-tier model above the detection band, so absence of signal there should be read as not measurable, not unbiased. Together with the ceiling-effect characterisation in arXiv:2606.26185, this work brackets the detectable region of position-bias measurement and makes the field central question askable in a verifiable form. Package, data, preregistration under MIT.

Hiroki Tamba

Independent researcher.

Position Bias Large Language Model Evaluation Permutation Position Random Rotations Ceiling

Jul 21, 2026cs.LG

Relative Positions Generalize, Absolute Positions Memorize: An Implicit-Bias Account of Length Generalization in Attention

Transformers with relative positional encodings often extrapolate to sequences longer than those seen during training, whereas transformers with learned absolute encodings typically do not. This is a robust empirical regularity, and the explanations offered for it so far are chiefly about expressivity, that is, about whether a length-generalizing solution exists. We give an optimization explanation. On a minimal fixed-offset retrieval task that isolates positional selection, the gap is governed by the implicit bias of the trained attention head: among the many solutions that fit short sequences, which one gradient descent actually selects. We prove that rotary encodings make the attention logit a function of relative offset alone, an exact equivariance, so whatever selection rule is learned at training lengths is reproduced verbatim at every longer length. Learned absolute encodings instead leave out-of-range positions unconstrained, and the trained head pins to a fixed absolute position inside the training range. We characterize the learned rotary rule as a low-rank ``carrier'' kernel aligned with the target offset, and we derive the resulting graceful accuracy decay as an attention-dilution law; both predictions are confirmed across seeds and offsets. A linear-attention control shows the mechanism is specific to softmax: without normalization, training selects a min-norm interpolant that does not extrapolate. The phenomenon, the equivariance, and the carrier all transfer to a multi-layer, multi-head transformer trained on a full-sequence length-generalization task. The account connects the implicit bias of attention, implicit bias for extrapolation in recurrent models, and the learning side of the RASP-L conjecture.

Subham Singh, Ashutosh Mishra, Subha Raut

Department of Mathematics and Statistics, Mississippi State University · Independent Researcher

Positional Encoding Relative Position Sequence Modeling Absolute Pose Position Bias Extrapolation

Jul 14, 2026cs.AI

Accuracy and Normalized Accuracy under Length Bias: Analysis, Guidelines, and a Bayesian Alternative

Multiple-choice benchmarks that rank candidate completions by conditional log-probability suffer from a length bias: because log-probabilities sum over tokens, longer answers tend to be penalized relative to shorter ones in practice. A common mitigation is to normalize scores by completion length, but we show empirically that this heuristic frequently over-corrects, introducing a bias toward longer answers instead. We first analyze these scoring rules, characterizing when standard and length-normalized accuracy are appropriate and how their length biases depend on the distribution of completion lengths. Motivated by this analysis, we introduce \emph{Bayesian accuracy}, a scoring rule that computes the posterior probability of each candidate under an explicit prior over answer length, thereby removing linear length effects. Bayesian accuracy is a drop-in replacement for likelihood-based multiple-choice evaluation, requires no additional forward passes, and consistently exhibits lower empirical length bias than both standard and length-normalized accuracy across benchmarks and few-shot settings.

Koen Oostermeijer

Aleph Alpha Research

Position Bias Multimodal Evaluation

Jul 9, 2026cs.CL

When the Judge Changes, So Does the Measurement: Auditing LLM-as-Judge Reliability

An LLM-as-judge score can move even when the candidate responses stay fixed, simply because the evaluator has changed. We treat this evaluator-replacement ambiguity as a measurement-validity problem. Across four judgment datasets, we compare two upgrade paths available in practice: scaling Qwen3 dense judges from 1.7B to 32B parameters and moving across MiniMax M2-M2.7 released APIs. The main pattern is that judge upgrades are not interchangeable: only Qwen3 1.7B to 4B gives a robust adjacent gain, while MiniMax adjacent releases do not. Stronger judges reduce but do not remove position and verbosity bias. Repeated-sample juries add little when errors are correlated. Structured debate can move decisions substantially, but without parser and fallback logs those shifts cannot be attributed to deliberation. We argue that LLM-as-judge reports should include dataset slices, bias probes, error-dependence estimates, and protocol audit trails.

Zongyou Yang, Yinghan Hou, Xiaokun Yang

Dyson School of Design Engineering, Imperial College London London, United Kingdom · Department of Electrical and Electronic Engineering, Imperial College London London, United Kingdom

Llm-As-A-Judge Judges Large Language Model Judges Model Auditing Position Bias Measures

Jul 5, 2026cs.IR

LBR: Towards Mitigating Length Bias in Large Language Models for Recommendation

Large language models (LLMs) have recently emerged as powerful backbones for recommender systems by reformulating recommendation as a token-level generation task. Despite their promise, we identify a pervasive yet underexplored issue:

\textit{Length Bias}

. Because items are represented by textual descriptions of varying lengths, LLM-based recommenders can be systematically biased in two ways. On the input side, longer item descriptions occupy more tokens in the context and thus receive disproportionately large aggregate attention mass during user preference modeling. On the output side, decoding based on summed autoregressive log-likelihood score inherently disfavors long items. Worse still, conventional length normalization can introduce an additional bias and even degrade recommendation performance. To address this problem, we propose

\textbf{LBR}

(

\textbf{L}

ength

\textbf{B}

ias

\textbf{R}

eduction), a lightweight and model-agnostic framework for mitigating length bias in LLM-based recommendation. LBR mitigates input-side bias via Length-Aware Attention Calibration, which incorporates a length-dependent offset into attention logits to neutralize attention skew. For the output side, LBR introduces Effective Information Length Normalization, replacing naive token count with an information-theoretic length surrogate derived from the branching structure of the prefix tree. Extensive experiments on three real-world Amazon datasets and two representative LLM-based recommenders demonstrate that LBR substantially alleviates length bias while consistently improving recommendation accuracy and fairness, with negligible additional training and inference overhead (with an average NDCG@5 gain of 16.82%). The code is available at https://github.com/Void-JackLee/LBR.

Hongchen Li, Bohao Wang, Jingbang Chen +5

State Key Laboratory of Blockchain and Data Security, Zhejiang University. · College of Computer Science and Technology, Zhejiang University. · Zhejiang University Hangzhou, Zhejiang, China +3

Recommendation Large Language Model Bias Large Language Model Reranking Position Bias Information-Theoretic Limits

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

Large Language Model Judges Position Bias Ranking Comparison Pool-Based Active Learning Covariates

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

Position Bias Incomplete

Jun 21, 2026cs.LG

On the Position Bias of On-Policy Distillation

On-Policy Distillation (OPD) improves the learning efficiency of standard reinforcement learning through dense, token-level supervision from teachers. In the standard KL objective of OPD, token-level losses are uniformly averaged, implying equal weights for all tokens. However, we discover that not all tokens are created equal: as student rollouts grow longer, they deviate further from the teacher's distribution, leading to degraded supervision quality at later positions. As a result, OPD using only the first 30% of tokens can perform comparably to using all tokens, whereas OPD using only the last 30% of tokens barely learns anything. In this work, we provide a principled understanding of this issue through the lens of constrained optimization. Based on these insights, we derive Importance-Weighted On-Policy Distillation (IW-OPD), in which the weight assigned to each token depends on the accumulated discrepancy between the student's and teacher's distributions, naturally upweighting earlier tokens and downweighting later ones with larger deviations. We show that IW-OPD converges significantly faster than OPD, with better learning efficiency, and achieves better final performance than standard OPD in both same-size and cross-scale settings, improving performance up to 6.9 points on AIME-2025.

Yan Xie, Sijie Zhu, Tiansheng Wen +2

Xidian University · Georgia Institute of Technology · Amazon AGI SF Lab

Efficient On-Policy Distillation Token-Level Entropy Position Bias

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

Large Language Model Evaluation Positive--Unlabeled Learning Position Bias Process-Level Supervision Quantifying

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

Large Language Model Bias Biases Model Judgments Position Bias Epistemic State Replication Reasoning Skills

Jun 3, 2026cs.CL

A Systematic Evaluation of Positional Bias in Multi-Video Summarization with MLLMs

Multimodal Large Language Models (MLLMs) are increasingly used for video understanding, yet their reliability under multi-video inputs remains poorly understood. We study positional bias in multi-video summarization, where the quality of a per-video summary can change with the video's input slot even when the underlying content is unchanged. We construct a benchmark from ActivityNet and News videos, covering Cooking, Domestic, Leisure, and News settings with two- and four-video inputs. We evaluate nine open-source and proprietary MLLMs and measure position effects with three complementary metrics: Coverage, Directional Positional Bias (DPB), and Middle-Edge Gap (MEG). Our results show that positional effects are domain- and model-dependent: signed directional bias can be small even when middle positions underperform, and increasing visual or generation budget does not uniformly remove the imbalance. We further analyze prompt-level mitigation methods. Together, the results show that multi-video summarization remains sensitive to input protocol and position, motivating more robust order-invariant multimodal systems.

Huangchen Xu, Yuan Wu, Yi Chang

School of Artificial Intelligence, Jilin University · 2Engineering Research Center of Knowledge-Driven Human-Machine Intelligence, Jilin University · 3International Center of Future Science, Jilin University

Video Multimodal Large Language Models Multimodal Large Language Models Position Bias Summarization

Jun 1, 2026cs.IR

Attention Calibration for Position-Fair Dense Retrieval

Dense retrieval compresses a passage into a single vector, but this compression is positionally skewed: early content dominates the embedding, and retrieval degrades when the relevant span appears later. Prior work proposed an inference-time method that counteracts this skew by equalizing the pooling token's attention across passage segments. However, (i) it redistributes attention at a fixed strength, (ii) it forces the pooling token's attention to itself to a fixed basket-level mass despite substantial variation across layers and architectures, and (iii) its effect on retrieval has not been evaluated. We introduce a strength coefficient that interpolates between uncalibrated and fully equalized attention, together with an efficient implementation that reduces peak calibration memory overhead from 5-7 GiB to under 1 MiB. Across three embedding models and two pooling schemes, moderate calibration provides a better retrieval trade-off than full equalization. We introduce a variant that preserves the pooling token's self-attention mass and redistributes only the remaining mass. On a position-aware retrieval benchmark spanning 10 languages and 31 domains, a configuration selected on English FineWeb-PosQ and transferred without tuning reduces position sensitivity in all 16 evaluated length-quartile, model, and retrieval-setting combinations, by up to 43% relative, while improving nDCG@10 by up to 4.8% relative and leaving general retrieval effectiveness on NanoBEIR essentially unchanged. Calibration runs at indexing time, adding no query-time latency. We release our code at github.com/impresso/fair-sentence-transformers

Andrianos Michail, Elias Schuhmacher, Juri Opitz +2

University of Zurich · Zurich, Switzerland

Retrieval Layer Position Bias

May 27, 2026cs.AI

Do LLMs Build World Models From Text? A Multilingual Diagnostic of Spatial Reasoning

Whether large language models (LLMs) construct internal spatial world models from pure-text descriptions remains contested, and whether such capabilities transfer across languages has not been systematically studied. We introduce MentalMap, a multilingual diagnostic benchmark with a six-level capability hierarchy (L0-L5) spanning atomic spatial facts to generative world-graph construction, together with four diagnostic axes probing frame of reference, reading-direction bias, reasoning-effort allocation, and hallucination. MentalMap is built from 100 ProcTHOR household scenes, covers eight typologically diverse languages plus a structured-text control, and contains 39 task families across 1,950 evaluation cells. Evaluating thirteen LLMs across scales and model families, we identify a universal L3 reasoning cliff: no model retains even half of its L0 performance on viewpoint reasoning once baseline atomic accuracy exceeds 40%. The cliff persists across languages, scales, and prompting strategies, while structured-output failures and reasoning patterns vary substantially across models. Human evaluation under the identical pure-text protocol reproduces the same failure pattern, suggesting that the bottleneck arises from text-only working memory constraints rather than being specific to current LLM architectures. Our findings reframe pure-text spatial reasoning as a multi-axis world-modeling problem and motivate multimodal and scratchpad-augmented reasoning as future directions.

Zhikai Pan, Chih-Ting Liao, Chunrui Liu +5

University of New South Wales · Essential Energy · University of Alabama at Birmingham +3

LLM Reasoning Strategies Spatial Reasoning Fuzzy Cognitive Map Text Analysis and Detection World Models Position Bias

May 27, 2026cs.AI

Examining Agents' Bias Amplification versus Suppression in Multi-Agent Systems

Multi-agent systems are increasingly deployed to support various tasks where agents interact to achieve individual and collective objectives. Although these systems can enhance task performance and decision-making, fairness preservation through bias reduction remains challenging. This study examines how agent-level biases shift and impact system-wide fairness. We use prompts to expose individual agents to group-favoring bias, then assess downstream impacts at the system level. To quantify the impact, we propose Favor Bias Strength (FBS), a zero-centered metric that decomposes bias alteration between favored-group uplift and disfavored-group suppression. Using multiple agent designs, benchmarks, and up-to-date large language models, we show that agents endowed with bias can substantially affect system-wide fairness. Interestingly, when agents are exposed to bias uniformly, the system-wide bias elevates, even exceeding the additive sum of the individual agents' biases. The empirical evidence underscores the criticality of fairness in multi-agent systems, which warrants further analyses and empirical tests.

Zejian Eric Wu, Zhongyi Jiang, Yuan Zhuang +1

Oregon State University · Independent Researcher · Amazon +1

Algorithmic Fairness Position Bias Suppression Objectives

May 26, 2026cs.CL

On the Hidden Costs of Counterfactual Knowledge Training in LLM Unlearning

Counterfactual tuning (CFT) has emerged as a promising paradigm for Large Language Model (LLM) unlearning by training models to generate alternative fictitious knowledge in place of undesired content. However, in this work, we find that this paradigm still underperforms other paradigms in some aspects, and identify two previously overlooked pitfalls underlying this gap: (1) knowledge conflict, where mutual inconsistencies within counterfactual corpora induce conflicting gradients that disrupt parameter optimization, and (2) hallucination spillover, where fitting false targets instills a persistent fabrication bias, inflating hallucination rates on unrelated domains. To systematically diagnose these issues, we introduce RWKU+, an extended benchmark equipped with novel trade-off metrics and gradient-level diagnostic tools. Our work further discusses the limitations and overhead of the paradigm, aiming to provide insights and actionable guidance for more rigorous LLM unlearning research.

Xiaotian Ye, Xiaohan Wang, Mengqi Zhang +1

Beijing University of Posts and Telecommunications · Huazhong University of Science and Technology · Shandong University +1

Large Language Model Unlearning Counterfactual Learning Large Language Models Fail Generative Hallucination Position Bias

May 23, 2026cs.CL

Who judges the judges? Governance from metrics: a runtime framework for continuous LLM compliance monitoring

Current approaches to AI compliance treat conformity as a binary, audit-time verdict rather than a continuous, measurable property of production systems. We argue that this compliance fiction is structurally ill-suited to the requirements of the EU AI Act, which demands ongoing human oversight and the detection of emergent behavioural drift in deployed systems. We introduce governance from metrics, a principle whereby regulatory compliance is derived as a continuous signal from runtime observability rather than from static assessments. Building on this principle, we present govllm, an open-source framework implementing a governance-driven routing architecture in which model selection is determined by accumulated compliance scores rather than by latency or cost alone. Central to our approach is a panel of regulatory judges - LLM evaluators specialised per criterion (EU AI Act, GDPR, ANSSI, accessibility) - whose inter-judge disagreement we reframe not as noise but as a regulatory uncertainty signal warranting human arbitration. We validate this approach through a ground truth corpus of 49 annotated prompt/response pairs across five regulatory criteria, evaluated by four small language models (SLMs, 1.7B-7B parameters) running fully on-premise. Agreement rates range from 51.5% (mistral:7b) to 69.1% (phi4-mini), with no single model dominating across all criteria - empirically motivating the Profile-as-jury design. We further document three structural failure modes in small regulatory judges and a judge-specific position bias that degrades agreement by up to 25 percentage points across three question-order conditions (original, reversed, permuted). govllm is released as open-source software to support reproducible AI governance research.

Jehanne Dussert

Independent Researcher

Compliance Artificial Intelligence Governance Llm-As-A-Judge Governance Large Language Model Evaluation Judges+3

May 20, 2026stat.ML

Semiparametric Efficient Bilevel Gradient Estimation

Functional bilevel methods estimate a lower-level function and plug it into a hypergradient, but this plug-in gradient can retain first-order bias when the lower-level problem is learned nonparametrically. To remove this bias, we develop a semiparametric debiasing theory for population bilevel gradients based on the efficient influence function. This perspective leads to a cross-fitted orthogonal hypergradient estimator for which we establish asymptotic normality together with uniform control over the outer parameter. Under quadratic losses, the estimator reduces to a simple doubly robust score based on conditional mean nuisances. On synthetic bilevel benchmarks with known ground truth, the method tracks the oracle efficient-gradient benchmark and improves over plug-in functional hypergradients and regularized kernel bilevel baselines.

Fares El Khoury, Houssam Zenati, Nathan Kallus +2

Université Grenoble Alpes, Inria, CNRS, Grenoble INP, LJK · Gatsby Computational Neuroscience Unit, University College London · Cornell University +1

Bilevel Optimization Influence Function Position Bias

May 13, 2026cs.CV

A $_3$ B $_2$ : Adaptive Asymmetric Adapter for Alleviating Branch Bias in Vision-Language Image Classification with Few-Shot Learning

Efficient transfer learning methods for large-scale vision-language models (

e.g.

, CLIP) enable strong few-shot transfer, yet existing adaptation methods follow a fixed fine-tuning paradigm that implicitly assumes a uniform importance of the image and text branches, which has not been systematically studied in image classification. Through extensive analysis, we reveal a Branch Bias issue in vision-language image classification: adapting the image encoder does not always improve performance under out-of-distribution settings. Motivated by this observation, we propose A

_3

B

_2

, an Adaptive Asymmetric Adapter that alleviates Branch Bias in few-shot learning. A

_3

B

_2

introduces Uncertainty-Aware Adapter Dampening (UAAD), which automatically suppresses image-branch adaptation when prediction uncertainty is high, enabling soft and data-driven control without manual intervention. Architecturally, A

_3

B

_2

adopts a lightweight asymmetric design inspired by mixture-of-experts with Load Balancing Regularization. Extensive experiments on three few-shot image classification tasks across 11 datasets demonstrate that A

_3

B

_2

consistently outperforms 11 competitive prompt- and adapter-based baselines.

Yiyun Zhou, Zhonghua Jiang, Wenkang Han +4

1Zhejiang University · 2Swansea University

Vision-Language Model Adaptation Few-Shot Learning Parameter-Efficient Adaptation Image Classification Position Bias

May 13, 2026cs.LG

Frequency Bias and OOD Generalization in Neural Operators under a Variable-Coefficient Wave Equation

Neural operators learn to map initial conditions to the terminal solution of partial differential equations (PDEs), providing a surrogate for the full operator mapping. This enables rapid prediction across different input configurations. While recent neural operator architectures have demonstrated strong performance on diverse PDE tasks, their behavior under structured distribution shifts remains insufficiently understood. To investigate this, we study operator learning in a wave propagation setting governed by a one-dimensional variable-coefficient wave equation, using two representative architectures, the Fourier Neural Operator (FNO) and the Deep Operator Network (DeepONet). To examine their generalization under distribution shifts, we consider structured out-of-distribution (OOD) settings that independently vary input frequency and coefficient smoothness. The results show that under smoothness shifts, both models maintain stable performance, with FNO achieving lower error. In contrast, under frequency shifts, FNO exhibits a sharp increase in error under unseen high-frequency inputs, whereas DeepONet shows milder degradation despite higher overall error. Our analysis reveals that these differences arise from how each architecture represents and responds to variations in frequency structure. Together, these findings highlight a fundamental gap between strong in-distribution performance and generalization under distribution shifts in operator learning, underscoring the role of architectural representation bias in developing more reliable neural operators for physics-based PDE simulations beyond the training distribution.

Runlong Xie, An Luo

Independent Researcher · School of Statistics, University of Minnesota, MN, USA.

Fourier Neural Operators Neural Operators Partial Differential Equations Out-Of-Distribution Distribution Shifts Position Bias

May 12, 2026cs.CL

DiffScore: Text Evaluation Beyond Autoregressive Likelihood

Autoregressive language models are widely used for text evaluation, however, their left-to-right factorization introduces positional bias, i.e., early tokens are scored with only leftward context, conflating architectural asymmetry with true text quality. We propose masked reconstruction as an alternative paradigm, where every token is scored using full bidirectional context. We introduce DiffScore, an evaluation framework built on Masked Large Diffusion Language Models. By measuring text recoverability across continuous masking rates, DiffScore eliminates positional bias and naturally establishes an evaluation hierarchy from local fluency to global coherence. We further provide diagnostic tools unavailable to autoregressive frameworks: multi-timestep quality profiles that decompose scores across masking rates, and bidirectional PMI decomposition that disentangles fluency from faithfulness. Experiments across ten benchmarks show that DiffScore consistently outperforms autoregressive baselines in both zero-shot and fine-tuned settings. The code is released at: https://github.com/wenlai-lavine/DiffScore.

Wen Lai, Yingli Shen, Dingnan Jin +4

1Ant Group · 2Tsinghua University · 3Technical University of Munich

Score-Based Diffusion Model Text Analysis and Detection Position Bias

May 12, 2026cs.CV

Logit-Attention Divergence: Mitigating Position Bias in Multi-Image Retrieval via Attention-Guided Calibration

Multimodal Large Language Models (MLLMs) have shown strong performance in multi-image cross-modal retrieval, yet suffer from severe position bias, where predictions are dominated by input order rather than semantic relevance. Through empirical analysis, we identify a phenomenon termed Logit-Attention Divergence, in which output logits are heavily biased while internal attention maps remain well-aligned with relevant visual evidence. This observation reveals a fundamental limitation of existing logit-level calibration methods such as PriDe. Based on this insight, we propose a training-free, attention-guided debiasing framework that leverages intrinsic attention signals for instance-level correction at inference time, requiring only a minimal calibration set with negligible computational overhead. Experiments on MS-COCO-based benchmarks show that our method substantially improves permutation invariance and achieves state-of-the-art performance, enhancing accuracy by over 40% compared to baselines. Code is available at https://github.com/brightXian/LAD.

Mingtao Xian, Yifeng Yang, Qinying Gu +2

Zhiyuan College, Shanghai Jiao Tong University, Shanghai, China · Shanghai Jiao Tong University, Shanghai, China · Shanghai Artificial Intelligence Laboratory, Shanghai, China +1

Multimodal Large Language Models Cross-Modal Attention Cross-Modal Position Bias

May 11, 2026cs.LG

Quotient-Categorical Representations for Bellman-Compatible Average-Reward Distributional Reinforcement Learning

Average-reward reinforcement learning requires estimating the gain and the bias, which is defined only up to an additive constant. This makes direct distributional analogues ill-posed on the real line. We introduce a quotient-space formulation in which state-indexed bias laws are identified up to a common translation, together with a categorical parameterization that respects this symmetry. On this quotient-categorical space, we define a projected average-reward distributional operator and show that it is well-defined, non-expansive in a coordinate Cramér metric, and admits fixed points. We then study sampled recursions whose mean-field maps are asynchronous relaxations of this operator. In an idealized centered-reward setting, a one-state temporal-difference update enjoys almost sure convergence together with finite-iteration residual bounds under both i.i.d. and Markovian sampling. When the gain is unknown, we augment the recursion with an online gain estimator, and prove non-expansiveness and Markovian convergence of the resulting coupled scheme. Finally, we show that synchronous exact updates are gain-independent at the quotient-law level, isolating a structural contrast between ideal quotient distributions and practical fixed-grid categorical representations.

Ege C. Kaya, Aliasghar Pourghani, Vijay Gupta +1

Elmore Family School of Electrical and Computer Engineering Purdue University West Lafayette, IN 47906

Distributional Reinforcement Learning Convergence Representation Space Position Bias

May 11, 2026cs.LG

Spurious Correlation Learning in Preference Optimization: Mechanisms, Consequences, and Mitigation via Tie Training

Preference learning methods like Direct Preference Optimization (DPO) are known to induce reliance on spurious correlations, leading to sycophancy and length bias in today's language models and potentially severe goal misgeneralization in future systems. In this work, we provide a unified theoretical analysis of this phenomenon, characterizing the mechanisms of spurious learning, its consequences on deployment, and a provable mitigation strategy. Focusing on log-linear policies, we show that standard preference-learning objectives induce reliance on spurious features at the population level through two channels: mean spurious bias and causal-spurious correlation leakage. We then show that this reliance creates an irreducible vulnerability to distribution shift: more data from the same training distribution fails to reduce the model's dependence on spurious features. To address this, we propose tie training, a data augmentation strategy using ties (equal-utility preference pairs) to introduce data-driven regularization. We demonstrate that this approach selectively reduces spurious learning without degrading causal learning. Finally, we validate our theory on log-linear models and provide empirical evidence that both the spurious learning mechanisms and the benefits of tie training persist for neural networks and large language models.

Christian Moya, Alex Semendinger, Guang Lin +1

Spurious Correlations Position Bias Implications

May 10, 2026cs.LG

Let the Target Select for Itself: Data Selection via Target-Aligned Paths

Targeted data selection seeks training examples from a candidate pool that improve downstream task performance. While trajectory-based selectors effectively guide this process, existing approaches typically construct reference states by warming up on the candidate pool itself, thereby inheriting pool-dependent distributional biases. To address this coupling, we introduce Target-Aligned Candidate Selection (TACS), which constructs a short, capacity-constrained reference path using a compact target-validation proxy. Candidates are ranked by their normalized loss reduction along this trajectory, requiring only forward passes and allowing the path to be reused across distinct pools for a fixed model-target pair. Empirically, TACS achieves competitive downstream performance across controlled logistic, vision, and NLP benchmarks, while slashing selection compute by up to 69% and candidate storage to less than 10 MB in large-scale instruction tuning.

Huitao Yang, Hengzhi He, Tung Sum Thomas Kwok +1

University of California, Los Angeles

Alignment-Aware Selection Adaptive Sampling Reweighting Distributional Alignment Target Domain Instruction Tuning+2

May 7, 2026cs.CL

The Cost of Context: Mitigating Textual Bias in Multimodal Retrieval-Augmented Generation

While Multimodal Large Language Models (MLLMs) are increasingly integrated with Retrieval-Augmented Generation (RAG) to mitigate hallucinations, the introduction of external documents can conceal severe failure modes at the instance level. We identify and formalize the phenomenon of recorruption, where the introduction of even perfectly accurate "oracle" context causes a capable model to abandon an initially correct prediction. Through a mechanistic diagnosis of internal attention matrices, we show that recorruption is driven by a two-fold attentional collapse: (1) visual blindness, characterized by the systemic suppression of visual attention mass (

M_{vis}

) and sharpness (

S_{vis}

), and (2) a structural positional bias that forces the model to prioritize boundary tokens over semantic relevance. Our analysis reveals an Illusion of Success, demonstrating that many seemingly correct RAG outcomes are merely positional coincidences where the model's textual copying bias happens to align with the ground-truth location. To address these vulnerabilities, we propose Bottleneck Attention Intervention for Recovery (BAIR), a parameter-free, inference-time framework that restores visual saliency and applies position-aware penalties to textual distractors. Across medical factuality, social fairness, and geospatial benchmarks, BAIR successfully restores multimodal grounding and improves diagnostic reliability without requiring model retraining or fine-tuning.

Hoin Jung, Xiaoqian Wang

Elmore Family School of Electrical and Computer Engineering Purdue University West Lafayette, IN 47907

Multimodal Retrieval Augmented Generation Multimodal Large Language Models Distractor Position Bias Adversarial Corruption

May 6, 2026cs.CL

RLearner-LLM: Balancing Logical Grounding and Fluency in Large Language Models via Hybrid Direct Preference Optimization

Direct Preference Optimization (DPO), the efficient alternative to PPO-based RLHF, falls short on knowledge-intensive generation: standard preference signals from human annotators or LLM judges exhibit a systematic verbosity bias that rewards fluency over logical correctness. This blindspot leaves a logical alignment gap -- SFT models reach NLI entailment of only 0.05-0.22 despite producing fluent text. We propose RLearner-LLM with Hybrid-DPO: an automated preference pipeline that fuses a DeBERTa-v3 NLI signal with a verifier LLM score, removing human annotation while overcoming the "alignment tax" of single-signal optimization. Evaluated across five academic domains (Biology, Medicine, Law) with three base architectures (LLaMA-2-13B, Qwen3-8B, Gemma 4 E4B-it), RLearner-LLM yields up to 6x NLI improvement over SFT, with NLI gains in 11 of 15 cells and consistent answer-coverage gains. On Gemma 4 E4B-it (4.5B effective params), Hybrid-DPO lifts NLI in four of five domains (+11.9% to +2.4x) with faster inference across all five, scaling down to compact base models without losing the alignment-tax mitigation. Our Qwen3-8B RLearner-LLM wins 95% of pairwise comparisons against its own SFT baseline; GPT-4o-mini in turn wins 95% against our concise output -- alongside the 69% win the same judge gives a verbose SFT over our DPO model, this replicates verbosity bias on a frontier comparator and motivates logic-aware metrics (NLI, ACR) over LLM-as-a-judge for knowledge-intensive generation.

Qiming Bao, Juho Leinonen, Paul Denny +1

University of Auckland · Aalto University

Large Language Model Alignment Large Language Models(Llms Natural Language Inference Backtranslation Augmented Direct Preference Optimization Fluency Position Bias+1

May 3, 2026cs.CL

Do Large Language Models Plan Answer Positions? Position Bias in Multiple-Choice Question Generation

Large language models (LLMs) are increasingly used to generate multiple-choice questions (MCQs), where correct answers should ideally be uniformly distributed across options. However, we observe that LLMs exhibit systematic position biases during generation. Through extensive experiments with 10 LLMs and 5 vision-language models (VLMs) on three MCQ generation tasks, we show that these biases are structured, with similar patterns emerging within model families. To investigate the underlying mechanisms, we conduct probing experiments and find that hidden representations in the question stem encode predictive signals of the correct answer position, suggesting that answer position may be implicitly planned during generation. Building on this insight, we apply activation steering to manipulate internal representations and influence answer position. Our results show that steering can partially control positional preferences and substantially shift answer position distributions. Our findings provide a practical framework for studying implicit positional planning in LLMs and highlight the importance of controllable generation for reliable MCQ construction and evaluation.

Xuemei Tang, Xufeng Duan, Zhenguang G. Cai

The Hong Kong Polytechnic University · The Chinese University of Hong Kong

Multiple-Choice Questions Position Bias Question Distribution Shifts

May 3, 2026cs.CV

Exploring Entropy-based Active Learning for Fair Brain Segmentation

Active learning (AL) has emerged as a crucial strategy for reducing the prohibitive costs associated with medical image segmentation. However, standard uncertainty-based AL methods typically focus on maximizing performance metrics, ignoring performance disparities or fairness across groups with sensitive attributes. While fair active learning has been explored in classification tasks, its intersection with medical image segmentation remains unaddressed. In this work, we introduced a fairness-aware active learning framework with a Weighted Entropy selection strategy that modulates uncertainty based on current group-specific performance estimates on the labeled set. To decouple true epistemic uncertainty from anatomical volume variances, we further utilized a masked, scaled entropy restricted to the region of interest. The framework was evaluated on synthetic T1-weighted brain MRIs with controlled left caudate bias in both strong and weak bias settings. A 3D U-Net was trained to segment the left caudate under several AL strategies, starting from both demographically balanced and strongly imbalanced initial labeled sets. Experiments demonstrated that our method markedly reduces performance disparities between groups compared to random sampling and standard uncertainty sampling. By prioritizing poorly segmented subgroups during the AL cycles, our method consistently achieved the highest equity-scaled performance and reduced the disparity metric by 75% (strong bias) and 86% (weak bias) relative to standard entropy at the final budget. Overall, this work is among the first studies on fair AL for medical image segmentation, offering an efficient strategy to train more equitable models in resource-constrained environments.

Ghazal Danaee, Mélanie Gaillochet, Christian Desrosiers +2

École de technologie supérieure, Montréal, QC, Canada · Polytechnique Montréal, Montréal, QC, Canada · Mila - Quebec AI Institute, Montréal, QC, Canada

Semi-Supervised Medical Image Segmentation Brain Tumor Segmentation Active Learning Algorithmic Fairness Position Bias

May 2, 2026eess.AS

Toward Fair Speech Technologies: A Comprehensive Survey of Bias and Fairness in Speech AI

Speech technologies are deployed in high-stakes settings, yet fairness concerns remain fragmented across tasks and disciplines. Existing surveys either adopt a general machine-learning perspective that overlooks speech-specific properties or focus on a single task, missing failure patterns shared across the speech domain. Synthesizing over 400 studies spanning generation and perception tasks and emerging speech-language models, this survey presents a unified framework that links formal fairness definitions to evaluation, diagnosis, and mitigation. We formalize seven fairness definitions adapted to the speech modality and organize the field's conceptual evolution through three paradigms: Robustness, Representation, and Governance. We then ground evaluation metrics in the mathematical cores of these definitions and offer a decision tree for metric selection. We diagnose bias sources along the speech processing pipeline, surfacing speech-specific mechanisms such as channel bias as a demographic proxy and annotation subjectivity in emotion labels. We systematize mitigation strategies across four intervention stages, mapping each to the diagnosed sources. Finally, we identify open challenges and propose directions for future research.

Yi-Cheng Lin, Yun-Shao Tsai, Kuan-Yu Chen +3

Graduate Institute of Communication Engineering, National Taiwan University, Taipei, Taiwan · University of Southern California, Los Angeles, CA, USA · NTU Artificial Intelligence Center of Research Excellence (NTU AI-CoRE), National Taiwan University, Taipei, Taiwan

Speaker Large Language Model Bias Algorithmic Fairness Position Bias Evaluation Metrics Emotion Recognition

Apr 29, 2026cs.CL

Option-Order Randomisation Reveals a Distributional Position Attractor in Prompted Sandbagging

A predecessor pilot (Cacioli, 2026) found that Llama-3-8B implements prompted sandbagging as positional collapse rather than answer avoidance. However, fixed option ordering in MMLU-Pro left open whether this reflected a model-level position-dominant policy or dataset-level distractor structure. This pre-registered follow-up (3 models, 2,000 MMLU-Pro items, 4 conditions, 24,000 primary trials) added cyclic option-order randomisation as the critical control. The pre-registered item-level same-letter diagnostic did not confirm deterministic position-tracking (same-letter rate 37.3%, below the 50% threshold). However, pre-specified supporting analyses revealed that the response-position distribution under sandbagging was highly stable under complete content rotation (Pearson r = 0.9994; Jensen-Shannon divergence = 0.027, compared to 0.386 between honest and sandbagging conditions). Accuracy spiked to 72.1% when the correct answer coincidentally occupied the preferred position E, and fell to 4.3% at position A. The data provide strong evidence for a soft distributional attractor: under sandbagging instruction, the model enters a low-entropy response-position basin centred on E/F/G that is highly stable and largely content-invariant at the aggregate level. Qwen-2.5-7B served as a negative control (non-compliant, no distributional shift). These results provide evidence, at the 7-9 billion parameter scale, that response-position entropy is a promising black-box behavioural signature of this sandbagging mode.

Jon-Paul Cacioli

Independent Researcher, Melbourne, Australia

Mmlu-Pro Position Bias Distractor Random Sabotage Entropy+1

Apr 28, 2026cs.CL

Below-Chance Blindness: Prompted Underperformance in Small LLMs Produces Positional Bias Rather than Answer Avoidance

Detecting sandbagging--the deliberate underperformance on capability evaluations--is an open problem in AI safety. We tested whether symptom validity testing (SVT) logic from clinical malingering detection could identify sandbagging through below-chance performance (BCB) on forced-choice items. In a pre-registered pilot at the 7-9 billion parameter instruction-tuned scale (3 models, 4 MMLU-Pro domains, 4 conditions, 500 items per cell, 24,000 total trials), the plausibility gate failed. Zero of 12 model-domain cells showed significant below-chance performance under sandbagging instruction. Exploratory analyses revealed three qualitatively distinct failure modes. Qwen-2.5-7B and Phi-3.5-mini largely ignored the sandbagging instruction, with 62-88% response identity with the honest baseline. Llama-3-8B complied substantially but implemented underperformance as a positional heuristic, collapsing its response distribution onto middle-alphabet options (E at 31.8%, F at 26.1%) regardless of where the correct answer fell. This produced accuracy boosts of up to 33 percentage points when the correct answer coincidentally occupied the model's preferred position. An explicit anti-task instruction ("pick the least likely answer") drove two of three models below chance, with accuracy as low as 0.024. The capability for answer-aware avoidance therefore exists but is not activated by "deliberately underperform." BCB did not fail as a logical marker of answer-aware avoidance. It was not observed in this regime because the model showing the largest behavioural shift exhibited behaviour consistent with a position-dominant response policy rather than content-aware answer avoidance. We propose that positional-distribution shift may be a more effective behavioural signature than below-chance accuracy for detecting prompted underperformance at this model scale.

Jon-Paul Cacioli

Independent Researcher, Melbourne, Australia

Sabotage Position Bias Depression

Apr 26, 2026cs.CL

Where to Place the Query? Unveiling and Mitigating Positional Bias in In-Context Learning for Diffusion LLMs via Decoding Dynamics

While In-Context Learning (ICL) is extensively studied in Autoregressive (AR) LLMs, its mechanism within Diffusion Large Language Models (dLLMs) remains largely unexplored. Unlike AR models restricted by unidirectional causal masking, dLLMs intrinsically utilize bidirectional attention, offering extensive spatial flexibility for query placement. Unfortunately, current practices conventionally inherit AR-style trailing-query templates, often overlooking the structural paradigm shift. This paper presents a comprehensive analysis unveiling that query position is actually a first-order variable in dLLMs. Through empirical decoupling, we demonstrate that positional variance impacts generation quality on par with example semantic quality. Internally, this positional sensitivity stems from a spatial ``Recency Effect'' in attention flow and task-dependent shifts in decoding trajectories. To mitigate this instability without ground-truth labels, we reveal that traditional single-step confidence (

C_{decoded}

) fails in dLLMs. Instead, we propose Average Confidence (

\overline{C}

), a novel metric tracking the iterative decoding process. By establishing the foundational spatial ICL baselines, we introduce Auto-ICL, a training-free adaptive routing strategy that dynamically optimizes query placement, robustly approaching oracle performance across heterogeneous reasoning and perception tasks.

Zhengheng Li, Panrui Li, Xuyang Liu +1

Southeast University

Diffusion Language Models In-Context Learning Position Bias Position

Apr 25, 2026cs.AI

Judging the Judges: A Systematic Evaluation of Bias Mitigation Strategies in LLM-as-a-Judge Pipelines

LLM-as-a-Judge has become the dominant paradigm for evaluating language model outputs, yet LLM judges exhibit systematic biases that compromise evaluation reliability. We present a comprehensive empirical study comparing nine debiasing strategies across five judge models from four provider families (Google, Anthropic, OpenAI, Meta), three benchmarks (MT-Bench n=400, LLMBar n=200, custom n=375), and four bias types. Our headline practical finding is that a mid-tier model with the right debiasing can outperform frontier judges at a fraction of the cost: Gemini 2.5 Flash with the Combined Budget strategy reaches the highest agreement of any configuration we tested (71.0%, kappa=0.549) at ~

0.001 per evaluation, about 15x cheaper than the best frontier setup (Claude Sonnet 4, 69.5%, ~

0.015). Other key findings: (1) Style bias is the dominant bias (0.10-0.76 across models, favoring markdown over plain prose), far exceeding position bias (<=0.04), yet is rarely studied. (2) Verbosity bias is heterogeneous when measured length-aware: Pro, Flash, and Llama prefer longer answers (+0.24 to +0.44), Claude prefers concise (-0.12), and GPT-4o is neutral (-0.04); on truncation controls all models correctly prefer the complete response (0.88-1.00 accuracy). (3) Debiasing helps multiple models: Claude S8 (+11.5 pp), Flash S8 (+7.5 pp), and Claude S5 (+7.3 pp) survive Holm-Bonferroni correction, with Flash S1 (+4.7 pp) and Llama S8 (+4.5 pp) also significant. We release our evaluation framework, the 375-pair controlled dataset, and per-instance cached results for all nine strategies.

Sadman Kabir Soumik

Independent Researcher

Llm-As-A-Judge Judges Position Bias

Apr 23, 2026cs.CL

The Coin Flip Judge? Reliability and Bias in LLM-as-a-Judge Evaluation

LLM-as-a-Judge is now widely used to rank model outputs, train reward models, and populate public leaderboards, but its run-to-run reliability remains under-characterized. We study repeated identical evaluations on 29 tasks spanning 10 categories using two OpenAI judge models (GPT-4o-mini and GPT-4.1-mini), with 50 pairwise trials and 50 pointwise trials per question, supplemented by temperature and prompt-sensitivity ablations. Across judges, pairwise preferences flip on average 13.6% of the time, with 28% of questions exceeding a 20% flip rate and one question reaching 56%. GPT-4o-mini also exhibits a significant first-position bias (72% A-majority, p = 0.024). At the same time, mean pointwise score gaps are small (0.19--0.36 on a 10-point scale) and not statistically significant in aggregate, producing a pairwise--pointwise gap: judges frequently choose a winner even when their own scalar scores provide little evidence of a meaningful quality difference. Beyond within-judge instability, cross-judge agreement is only 76% (

κ= 0.51

), semantically equivalent prompt templates change majority outcomes in 25% of tested cases, and deterministic decoding reduces but does not eliminate inconsistency. A reliability curve analysis shows that, in our dataset, 11 repeated trials are needed for a majority vote to recover the 50-trial reference verdict with 95% probability on average, rising to 15 for high-variance questions. These findings suggest that single-trial LLM judging is often too noisy for high-stakes evaluation, and that multi-trial aggregation, position randomization, and explicit uncertainty reporting should be standard practice. Because both judges are from a single provider, cross-provider replication remains an important next step.

Abel Yagubyan

Independent Researcher

Llm-As-A-Judge Large Language Model Judges Flips Position Bias

Apr 21, 2026cs.LG

Debiased neural operators for estimating functionals

Neural operators are widely used to approximate solution maps of complex physical systems. In many applications, however, the goal is not to recover the full solution trajectory, but to summarize the solution trajectory via a scalar target quantity (e.g., a functional such as time spent in a target range, time above a threshold, accumulated cost, or total energy). In this paper, we introduce DOPE (debiased neural operator): a semiparametric estimator for such target quantities of solution trajectories obtained from neural operators. DOPE is broadly applicable to settings with both partial and irregular observations and can be combined with arbitrary neural operator architectures. We make three main contributions. (1) We show that, in contrast to DOPE, naive plug-in estimation can suffer from first-order bias. (2) To address this, we derive a novel one-step, Neyman-orthogonal estimator that treats the neural operator as a high-dimensional nuisance mapping between function spaces, and removes the leading bias term. For this, DOPE uses a weighting mechanism that simultaneously accounts for irregular observation designs and for how sensitive the target quantity is to perturbations of the underlying trajectory. (3) To learn the weights, we extend automatic debiased machine learning to operator-valued nuisances via Riesz regression. We demonstrate the benefits of DOPE across various numerical experiments.

Konstantin Hess, Dennis Frauen, Niki Kilbertus +1

1LMU Munich · 2Munich Center for Machine Learning · 3TU Munich +1

Neural Operators Neural Approximations Single Trajectory Position Bias

Apr 21, 2026cs.AI

More Thinking, More Bias: Length-Driven Position Bias in Reasoning Models

Chain-of-thought (CoT) reasoning and reasoning-tuned models such as DeepSeek-R1 are commonly assumed to reduce shallow heuristic biases by thinking carefully. We test this on position bias in multiple-choice QA and find a different story: within any reasoning-capable model, per-question position bias scales with the length of the reasoning trajectory. Across thirteen reasoning-mode configurations (two R1-distilled 7-8B models, two base models prompted with CoT, and DeepSeek-R1 at 671B) on MMLU, ARC-Challenge, and GPQA, twelve show a positive partial correlation between trajectory length and Position Bias Score (PBS) after controlling for accuracy, ranging from 0.11 to 0.41 (all p < 0.05). All twelve open-weight reasoning-mode configurations show monotonically increasing PBS across length quartiles. A truncation intervention provides causal evidence: continuations resumed from later points in the trajectory are increasingly likely to shift toward position-preferred options (16% to 32% for R1-Qwen-7B across absolute-position buckets). At 671B, aggregate PBS collapses to 0.019, but the length effect still manifests in the longest quartile (PBS = 0.071), suggesting that accuracy gates the expression of length-driven bias rather than eliminating the underlying mechanism. We additionally find that direct-answer position bias is a distinct phenomenon with a different footprint (strong in Llama-Instruct-direct, weak in Qwen-Instruct-direct, and uncorrelated with trajectory length): CoT reasoning replaces this baseline bias with length-accumulated bias. Our results argue that reasoning-capable models should not be treated as order-robust by default in MCQ evaluation pipelines, and offer a diagnostic toolkit (PBS, commitment change point, effective switching, truncation probes) for auditing position bias in reasoning models.

Xiao Wang

Large Reasoning Models Position Bias Chain-of-Thought Reasoning Evaluation Agent

Apr 18, 2026cs.AI

Machine individuality: Separating genuine idiosyncrasy from response bias in large language models

As large language models (LLMs) are increasingly integrated into daily life, in roles ranging from high-stakes decision support to companionship, understanding their behavioral dispositions becomes critical. A growing literature uses psychometric inventories and cognitive paradigms to profile LLM dispositions. However, these approaches cannot determine whether behavioral differences reflect stable, stimulus-specific individuality or global response biases and stochastic noise. Here, we apply crossed random-effects models -- widely used in psychometrics to separate systematic effects -- to 74.9 million ratings provided by 10 open-weight LLMs for over 100,000 words across 14 psycholinguistic norms. On average, 16.9% of variance is attributable to stimulus-specific individuality, robustly exceeding a statistical null model. Cross-norm prediction analyses reveal this individuality as a coherent fingerprint, unique to each model. These results identify individual differences among LLMs that cannot be attributed to response biases or stochastic noise. We term these differences machine individuality.

Valentin Kriegmair, Dirk U. Wulff

Center for Adaptive Rationality, Max Planck Institute for Human Development, 14195 Berlin, Germany · Center for Cognitive and Decision Sciences, Department of Psychology, University of Basel, 4055 Basel, Switzerland

Large Language Model Bias Linguistics Biases Position Bias Generalized Linear Model

Feb 2, 2026cs.CL

Am I More Pointwise or Pairwise? Revealing Position Bias in Rubric-Based LLM-as-a-Judge

Large language models are widely employed as evaluators, a paradigm commonly referred to as LLM-as-a-judge. Prior research has predominantly examined point-wise or pair-wise evaluation protocols; in contrast, our focus is on rubric-based evaluation, which has been attracting increasing attention owing to its utility for training models in domains where verification is otherwise difficult. In this work, we show that rubric-based evaluation implicitly resembles a multiple-choice setting and therefore exhibits position bias: LLMs tend to prefer score options that appear at specific positions within the rubric list. Through controlled experiments across multiple models and datasets, we demonstrate that this position bias is consistent. Its direction, however, is model-specific: some judges favor the first option, while others favor the last. We further identify a second, orthogonal axis of bias: when a prompt scores several criteria simultaneously, the ordering of the criteria itself shifts the resulting scores. We additionally explore permuting the order of the rubric options as a means of mitigating position bias, and find that although the bias can be attenuated, improvements in the correlation between model judgments and human annotations are obtained primarily for models that exhibit strong bias. Our results recast rubric-based LLM-as-a-judge as a multiple-choice problem with measurable, model-specific position bias, and we further confirm that only a small number of random order permutations are sufficient to reduce the error introduced by this bias for the majority of models.

Yuzheng Xu, Tosho Hirasawa, Tadashi Kozuno +1

OMRON SINIC X · NexaScience

Llm-As-A-Judge Model Judgments Position Bias Grading

Oct 1, 2025cs.SD

Hearing the Order: Investigating Position Bias in Large Audio-Language Models

Large audio-language models (LALMs) are often used in tasks that involve reasoning over ordered options. An open question is whether their predictions are influenced by the order of answer choices, which would indicate a form of position bias and undermine their reliability. In this paper, we identify and analyze this problem in LALMs. We demonstrate that no model is immune to this bias through extensive experiments on six LALMs across three widely used benchmarks and their spoken counterparts. Shuffling the order of answer options can cause performance fluctuations of up to 24% and even change model rankings, raising concerns about the reliability of current evaluation practices. We also study permutation-based strategies and show that they can mitigate bias in most cases. Our work represents the first systematic investigation of this issue in LALMs, and we hope it raises awareness and motivates further research in this direction.

Yu-Xiang Lin, Chen-An Li, Sheng-Lun Wei +3

National Taiwan University

Large Audio Language Models Large Language Model Bias Position Bias Order Permutation Answer+1

May 19, 2025cs.LG

Bias Fitting to Mitigate Length Bias of Reward Model in RLHF

Reinforcement Learning from Human Feedback (RLHF) relies on reward models to align large language models with human preferences. However, RLHF often suffers from reward hacking, wherein policy learning exploits flaws in the trained reward model to maximize reward scores without genuinely aligning with human preferences. A significant example of such reward hacking is length bias, where reward models usually favor longer responses irrespective of actual response quality. Previous works on tackling length bias have notable limitations, these approaches either mitigate bias without characterizing the bias form, or simply assume a linear length-reward relation. To accurately model the intricate nature of length bias and facilitate more effective bias mitigation, we propose FiMi-RM (Bias Fitting to Mitigate Length Bias of Reward Model), a framework that autonomously learns and corrects underlying bias patterns. Our approach consists of three stages: First, we warm up by training a standard reward model which inherently contains length bias. Next, we deploy a lightweight fitting model to capture the non-linear relation between length and reward. Finally, we incorporate this learned relation into the reward model, effectively decoupling length from reward while preserving preference modeling capabilities. Experimental results demonstrate that FiMi-RM achieves a more balanced length-reward distribution. Furthermore, when applied to alignment algorithms such as Direct Preference Optimization (DPO) and Best-of-N (BoN), our debiased reward model improves length-controlled win rate and reduces verbosity without compromising its performance.

Kangwen Zhao, Jianfeng Cai, Jinhua Zhu +5

University of Science and Technology of China

Reinforcement Learning From Human Feedback Large Language Model Alignment Position Bias