Overview

This field examines how automated judges can favor certain answers for reasons unrelated to quality, including response order, writing style, verbosity, model identity, language, and demographic or cultural cues. Researchers use controlled counterfactual tests and benchmarks such as MT-Bench, RewardBench, and multilingual or domain-specific evaluation sets to measure when judgments change and whether they match human ratings. Mitigation approaches include randomizing presentation, adding rubrics and reference answers, calibrating scores with human comparisons, combining judges robustly, and training models to separate relevant evidence from distracting cues. A central challenge is that repeatable judgments and agreement among models do not guarantee validity, while interventions that reduce bias can also erase real quality differences or legitimate differences in human preferences.

Latest papers

128 papers11 papers in the last four weeks, down 56% on the four weeks before. 0.1% of all new papers.

Feb 14, 2026

Rubrics as an Attack Surface: Stealthy Preference Drift in LLM Judges

Evaluation and alignment pipelines for large language models increasingly rely on LLM-based judges, whose behavior is guided by natural-language rubrics and validated on benchmarks. We identify a previously under-recognized vulnerability in this workflow, which we term Rubric-Induced Preference Drift (RIPD). Even when rubric edits pass benchmark validation, they can still produce systematic and directional shifts in a judge's preferences on target domains. Because rubrics serve as a high-level decision interface, such drift can emerge from seemingly natural, criterion-preserving edits and remain difficult to detect through aggregate benchmark metrics or limited spot-checking. We further show this vulnerability can be exploited through rubric-based preference attacks, in which benchmark-compliant rubric edits steer judgments away from a fixed human or trusted reference on target domains, systematically inducing RIPD and reducing target-domain accuracy up to 9.5% (helpfulness) and 27.9% (harmlessness). When these judgments are used to generate preference labels for downstream post-training, the induced bias propagates through alignment pipelines and becomes internalized in trained policies. This leads to persistent and systematic drift in model behavior. Overall, our findings highlight evaluation rubrics as a sensitive and manipulable control interface, revealing a system-level alignment risk that extends beyond evaluator reliability alone. The code is available at: https://github.com/ZDCSlab/Rubrics-as-an-Attack-Surface. Warning: Certain sections may contain potentially harmful content that may not be appropriate for all readers.
Feb 6, 2026

FairJudge: An Adaptive, Debiased, and Consistent LLM-as-a-Judge

Existing LLM-as-a-Judge systems suffer from three fundamental limitations: limited adaptivity to task- and domain-specific evaluation criteria, systematic biases driven by non-semantic cues such as position, length, format, and model provenance, and evaluation inconsistency that leads to contradictory judgments across different evaluation modes (e.g., pointwise versus pairwise). To address these issues, we propose FairJudge, an adaptive, debiased, and consistent LLM-as-a-Judge. Unlike prior approaches that treat the judge as a static evaluator, FairJudge models judging behavior itself as a learnable and regularized policy. From a data-centric perspective, we construct a high-information-density judging dataset that explicitly injects supervision signals aligned with evaluation behavior. Building on this dataset, we adopt a curriculum-style SFT-DPO-GRPO training paradigm that progressively aligns rubric adherence, bias mitigation, and cross-mode consistency, while avoiding catastrophic forgetting. Experimental results on multiple internal and public benchmarks show that FairJudge consistently improves agreement and F1, reduces non-semantic biases, and outperforms substantially larger instruction-tuned LLMs. All resources will be publicly released after acceptance to facilitate future research.
Feb 2, 2026

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.
Feb 1, 2026

Agents-as-Peers: Bias-Controlled LLM Evaluation as Multi-Agent Peer Assessment

Large language model (LLM) evaluation commonly relies on fixed human-authored benchmarks or centralized human/model judges. Fixed benchmarks can age and enter model training corpora, while centralized judges introduce their own scoring preferences. We cast LLM evaluation as a multi-agent peer-assessment problem and introduce Agents-as-Peers, a distributed evaluation protocol in which every agent generates tasks, answers the shared task set, and judges the answers of its peers, with no privileged evaluator. Self-ratings are excluded from the ranking, and judging is repeated under controlled identity and presentation regimes so that self bias, identity effects, presentation-position sensitivity, and judge severity are measured rather than left implicit. In an instantiation with 12 LLMs and 420 agent-generated questions, Agents-as-Peers aggregates 55,440 non-self peer judgments per regime. Most agents assign their own answers higher scores than those answers receive from peer judges; visible identity changes scores for some agents; the first fixed answer slot is associated with scores 0.39 points higher than the shuffled baseline; and judges differ markedly in severity. On labeled benchmarks, peer scores align strongly with accuracy on TruthfulQA (Pearson r=0.904) and GSM8K (r=0.873). On TruthfulQA, the peer-score correlation is substantially higher than the self-score correlation (r=0.538). A benchmark-overlap screen and item-level human audit further characterize the endogenous task set. These results support Agents-as-Peers as a cohort-relative measurement protocol whose ranking is reported together with the biases and rater effects that shape it. Code and dataset: https://github.com/caura-ai/caura-PeerRank
Jan 30, 2026

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

Recent research has shown that large language models (LLMs) favor their own outputs when acting as judges, undermining the integrity of automated post-training and evaluation workflows. However, it is difficult to disentangle which behaviors are explained by narcissism versus experimental confounds. Specifically, LLM evaluators may deliver self-preferring verdicts when comparing responses to questions they fail on; these verdicts may not depend on the identity of the author, but on evaluator quality. We correct this by directly comparing the judge's voting distribution in cases where it evaluates itself versus another model. This evaluator quality baseline reveals that only 51% of examples in previous findings retain statistical significance against this null hypothesis, covering 89.6% of total self-preference probability mass. Finally, we compare the entropy of voting distributions, suggesting uncertainty-driven overlap, and show that our procedure enables more careful documentation against the backdrop of judge-bias research.
Sep 30, 2025

When LLMs Benchmark Themselves: Deconstructing Self-Bias in Automated Evaluation

As LLMs rapidly saturate existing benchmarks, automated benchmark creation using LLMs (LLM as a benchmark) where a model generates test inputs (LLM as a testset) and evaluates outputs (LLM as an evaluator) has gained traction as a cheap alternative to human curation. We show that this paradigm has a fundamental problem: LLM-generated benchmarks systematically favor the model that created them. Using machine translation as our primary testbed, we find that self bias arises from two compounding sources, LLM as a testset and LLM as an evaluator, and their combination amplifies the effect. Crucially, even when test data is generated with explicit diversity controls, each modelś implicit stylistic tendencies produce homogeneous, model-specific outputs that inflate its own scores. Increasing source text diversity, using our proposed diversity metric, partially mitigates this bias. Self bias is strong enough to cause each model to rank itself first, overriding the peer consensus ordering. We confirm that the phenomenon extends to open-ended generation on the Chatbot Arena task.
Sep 3, 2025

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

Large language models (LLMs) increasingly serve as automated evaluators, yet they suffer from "self-preference bias": a tendency to favor their own outputs over those of other models. This bias undermines fairness and reliability in evaluation pipelines, particularly for tasks like preference tuning and model routing. We investigate whether lightweight steering vectors can mitigate this problem at inference time without retraining. We introduce a curated dataset that distinguishes self-preference bias into justified examples of self-preference and unjustified examples of self-preference, and we construct steering vectors using two methods: Contrastive Activation Addition (CAA) and an optimization-based approach. Our results show that steering vectors can reduce unjustified self-preference bias by up to 97%, substantially outperforming prompting and direct preference optimization baselines. Yet steering vectors are unstable on legitimate self-preference and unbiased agreement, implying self-preference spans multiple or nonlinear directions. This underscores both their promise and limits as safeguards for LLM-as-judges and motivates more robust interventions.
Jul 15, 2025

Let's Think in Two Steps: Mitigating Agreement Bias in MLLMs with Self-Grounded Verification

Verifiers--functions assigning rewards to agent behavior--have been key to AI progress in math, code, and games. However, extending gains to domains without clear-cut success criteria (e.g., computer use) remains a challenge: while humans can recognize desired outcomes, translating this intuition into scalable rules is nontrivial. Multimodal LLMs (MLLMs) offer a promising solution, given their world knowledge, human-preference alignment, and reasoning skills. We evaluate MLLM verifiers across web navigation, computer use, and robotics, spanning 13 models, 28 judge designs, diverse agent trajectories, and distinct downstream applications. We identify a critical limitation: a strong tendency for MLLMs to over-validate agent behavior, a phenomenon we term agreement bias. This bias is pervasive across models, is resilient to test-time scaling, and harms pipelines relying on MLLM judgments (e.g., self-improvement). We discuss principles for design and evaluation of verifiers, and introduce SGV, a method that improves verification by harnessing MLLMs' own capabilities via modulation of (un)conditional generation. First, an MLLM generates broad priors about desired behavior, independent of the data under evaluation. Then, conditioned on generated priors, it reasons over and judges a candidate trajectory. Our methods yield more human-aligned verifiers, with gains of up to 25pp in failure detection, 14pp in accuracy, and benefits extending to downstream applications. In self-improvement and online supervision, our verifiers boost task completion of a GUI specialist in OSWorld, a diffusion policy in robomimic, and a ReAct agent in VisualWebArena--surpassing the previous SOTA by 20pp. Finally, we release an updated version of VisualWebArena, featuring improved evaluators, high-fidelity environment parallelism with >10x speedups, and VWA-Lite, a 1/3-scale subset with comparable evaluation fidelity.