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.

Large language models (LLMs) are widely used as reference-free evaluators via prompting, but this "LLM-as-a-Judge" paradigm is costly, opaque, and sensitive to prompt design. In this work, we investigate whether smaller models can serve as efficient evaluators by leveraging internal representations instead of surface generation. We uncover a consistent empirical pattern: small LMs, despite with weak generative ability, encode rich evaluative signals in their hidden states. This motivates us to propose the Semantic Capacity Asymmetry Hypothesis: evaluation requires significantly less semantic capacity than generation and can be grounded in intermediate representations, suggesting that evaluation does not necessarily need to rely on large-scale generative models but can instead leverage latent features from smaller ones. Our findings motivate a paradigm shift from LLM-as-a-Judge to Representation-as-a-Judge, a decoding-free evaluation strategy that probes internal model structure rather than relying on prompted output. We instantiate this paradigm through INSPECTOR, a probing-based framework that predicts aspect-level evaluation scores from small model representations. Experiments on reasoning benchmarks (GSM8K, MATH, GPQA) show that INSPECTOR substantially outperforms prompting-based small LMs and closely approximates full LLM judges, while offering a more efficient, reliable, and interpretable alternative for scalable evaluation. The code and data are available at: https://github.com/zhuochunli/Representation-as-a-judge

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.

Users increasingly rely on Large Language Models (LLMs) for Deep Research, using them to synthesize diverse sources into structured reports that support understanding and action. In this context, the practical reliability of such reports hinges on logical quality: whether the report's claims and arguments are explicitly supported and can be trusted as a basis for downstream use, rather than merely appearing fluent or informative. However, current evaluation frameworks largely overlook this requirement. To bridge this gap, we introduce ReportLogic, a benchmark that quantifies report-level logical quality through a reader-centric lens of auditability. Specifically, ReportLogic adopts a hierarchical taxonomy that evaluates whether readers can (1) trace an on-topic report structure with a unified analytical arc (Macro-Logic), (2) understand the progression with necessary context (Expositional-Logic), and (3) verify conclusions via explicit claim--support (Structural-Logic). Based on this taxonomy, we construct a human-annotated rubric-guided dataset and train an open-source LogicJudge for scalable evaluation. We further evaluate judge robustness via adversarial attacks, showing that off-the-shelf LLM judges are frequently influenced by superficial cues (e.g., verbosity), and reasoning modes can mask broken support relations. Overall, our results provide actionable guidance for building more robust logic evaluators and improving the logical reliability of LLM-generated reports.

Rubric-based text evaluation increasingly relies on large language models (LLMs) as scalable judges, yet frozen black-box models can interpret the same criteria inconsistently, produce score attributions that are difficult to audit, and map judgments poorly onto human scoring scales. We define this challenge as criteria transfer: translating human rubric intent into a stable, auditable inference-time scoring protocol. We introduce Rulers, which locks a task-level rubric specification, executes it through structured, evidence-grounded judgments, and calibrates the resulting signals to human score boundaries. Across four rubric-governed benchmarks and multiple frozen backbone models, Rulers achieves stronger agreement with human scores in most evaluated settings, while better matching empirical score distributions and remaining more stable under semantically equivalent rubric perturbations. Calibration controls and component ablations show that these gains cannot be attributed to post-hoc alignment alone, but depend on the combination of fixed criteria, traceable evidence, and calibrated score interpretation. These findings suggest that reliable LLM judging requires faithfully operationalizing human evaluation standards rather than relying on prompt-level scoring alone. Our code is available at https://github.com/LabRAI/Rulers.git.

As Natural Language Generation (NLG) dominates modern NLP, scalable evaluation remains a critical bottleneck. Consequently, LLM-as-a-judge (LaaJ) adoption has accelerated rapidly, appearing in more papers than human evaluation in 2025. This pivotal shift motivates a critical analysis of current evaluation practices. Overcoming the limits of rigid keyword filtering and manual review, we employ a multi-LLM information extraction pipeline to gather structured metadata from 14,171 papers across four major NLP conferences (2020-2025). Analyzing 3,334 filtered NLG papers, we identify three systemic challenges. (1) Metric inertia: despite the shift toward open-ended generation, legacy lexical metrics (BLEU, ROUGE) persist as primary indicators, typically used alongside rather than replaced by semantic alternatives. (2) Metric-criteria mapping problem: our paper-level co-occurrence data reveals that general-purpose automatic metrics are applied as broad proxies for quality, without specifying which dimension of text generation they are intended to evaluate. (3) Validation gap: LaaJ has grown rapidly without commensurate human validation (fewer than 8% of papers). Crucially, while LaaJ correlates with aggregate quality, alignment collapses on fine-grained criteria like fluency. To address these gaps, we distill our findings into a minimal Evaluation Checklist to guide metric selection, construct validity, and LaaJ deployment.

Large language models (LLMs) now support contexts of up to 1M tokens, but their strengths and weaknesses on complex long-context tasks remain unclear. To study this, we focus on multi-document legal case summarization, where a single case often spans many documents exceeding 100K tokens. We systematically evaluate 12 frontier LLMs with Gavel, which consists of Gavel-Ref, a reference-based evaluation framework with checklist, residual-fact, and writing-style evaluations, and Gavel-Agent, a reference-free agent for evaluating factual coverage directly from source documents. Our results show that current models are more prone to omitting key information than hallucinating. They all perform well on simple checklist items, such as filing date, but struggle with rare and complex items, such as settlements. Performance also declines as case length increases. To meta-evaluate Gavel, we collect 160 hours of human annotations. Gavel-Agent reduces token usage by at least 36% compared to end-to-end and chunk-by-chunk methods while achieving competitive performance. Gavel-Agent also generalizes to the medical domain, performing the best with at least 77% fewer tokens.

Code from LLM-assisted development may contain defects like API hallucinations, motivating the development of code evaluation metrics to assess such code before deployment. As conventional rule-based metrics merely score code based on similarity to reference programs, and execution-based metrics require expensive manual preparation for tests, researchers have developed LLM-as-a-judge metrics for both behavioral and nonbehavioral code evaluation, and curated multiple benchmarks to validate their effectiveness. However, such benchmarks suffer from critical limitations: Some feature coarse-grained binary labels of passing/failing unit tests, unable to distinguish between fundamental and minor flaws; others utilize fine-grained criteria with manually annotated scores as ground-truth, but such criteria are often vaguely defined, introducing subjectivity and unreliability in human scores. Furthermore, they often directly prompt LLMs to generate code for evaluation, which may lead to unbalanced code quality distributions, failing to test the metrics' discriminative capability across varying quality levels. To curate a diverse benchmark with a balanced program quality distribution and streamline manual annotation, we propose AXIOM, a novel perturbation-based framework. It reframes program scores as the refinement effort needed for deployment, with two stages: (1) Quality-constrained semantic perturbation, prompting LLMs to apply sequences of predefined perturbation rules to existing programs, precisely controlling each program's target quality to achieve a balanced distribution. (2) Hybrid quality calibration, selecting semantically distinct programs for benchmark diversity, and performing streamlined manual annotation with multisource context. With AXIOM, we curate AXIOMBENCH, a large-scale multilingual benchmark with 1,962 code evaluation problems in a more balanced score distribution...

LLM-as-a-Judge systems supply the reward signal in modern RLHF and RLVR pipelines, but their binary verdict reduces to a single linear readout F_gap on one hidden state. We show this readout is shallow enough that short, low-perplexity tokens flip the verdict from "No" to "Yes". These tokens are sampled from the judge's own next-token distribution at the response position, with no manual seed set and no gradient-based optimization. Our procedure, AdvJudge-Zero, reaches $>$90% ensemble false-positive rate on 22 of 24 (model, dataset) cells across six Qwen, Llama, and Gemma judges, versus 54-72% for the prior curated 10-token benchmark, and the discovered surface transfers cross-format to a 70B scalar reward model. The same discovered pool enables a defense: a LoRA fine-tune stratified by a 9-class mechanism taxonomy hardens cross-family generalization where naive sampling on the same pool fails, with mechanism breadth rather than pool size carrying the gain. Under GRPO training, the hardened judge eliminates the reward-collapse failures (false-positive spikes and length collapse) we observe in the unhardened baseline on both MATH and GSM8K at ten seeds per condition. The discovered pool, the mechanism taxonomy, and per-prompt flip records will be released under responsible disclosure.

Reliable assessment of the abilities of large audio language models (LALMs) is essential to advancing the state of the art. As benchmarks rapidly evolve to incorporate complex reasoning and subjective tasks, they increasingly necessitate open-ended responses from LALMs. We present Open-ended Response Correctness Assessment (ORCA) -- a reliable and lightweight model-based approach for answer correctness and disagreement modeling. We employ a three-stage annotation pipeline combining human judgment, structured feedback, and human-AI correction, yielding 9,663 annotations across 3,699 question-answer pairs from 15 LALMs on three audio understanding and reasoning benchmarks (achieving a Krippendorff's alpha of 0.82). Our experiments employing curriculum learning show that ORCA models achieve a Spearman correlation of 0.91 with average human correctness ratings on seen benchmarks and generalize to unseen benchmarks with a score of 0.85, outperforming several LLM judge baselines including Gemini 2.5 Flash. Furthermore, we demonstrate that ORCA's predicted variance correlates strongly with human disagreement, allowing it to effectively identify problematic benchmark items.

Large language models (LLMs) have demonstrated significant potential to accelerate scientific discovery as valuable tools for analyzing data, generating hypotheses, and supporting innovative approaches in various scientific fields. In this work, we investigate how LLMs can handle the transition from conceptual research ideas to well-structured research plans. Effective research planning not only supports scientists in advancing their research but also represents a crucial capability for the development of autonomous research agents. Despite its importance, the field lacks a systematic understanding of LLMs' research planning capability. To rigorously measure this capability, we introduce the Idea2Plan task and Idea2Plan Bench, a set of benchmarks built from ICML 2025 and Nature Mental Health papers released after major LLM training cutoffs. Each benchmark instance includes a research idea and a grading rubric capturing the key components of valid plans. We further propose Idea2Plan JudgeEval, a complementary benchmark to assess the reliability of LLM-based judges against expert annotations. Experimental results show that GPT-5 achieves the strongest performance on the benchmark, though substantial headroom remains for improvement. Our study provides new insights into LLMs' capability for research planning and lays the groundwork for future progress.

Large language models (LLMs) are rapidly being adopted for tasks like drafting emails, summarizing meetings, and answering health questions. In these settings, users may need to share private information (e.g., contact details, health records). To evaluate LLMs' ability to identify and redact such information, prior work introduced real-life, scenario-based benchmarks (e.g., ConfAIde, PrivacyLens) and found that LLMs can leak private information in complex scenarios. However, these evaluations relied on proxy LLMs to judge the helpfulness and privacy-preservation quality of LLM responses, rather than directly measuring users' perceptions. To understand how users perceive the helpfulness and privacy-preservation quality of LLM responses to privacy-sensitive scenarios, we conducted a user study ($n=94$) using 90 PrivacyLens scenarios. We found that users had low agreement with each other when evaluating identical LLM responses. In contrast, five proxy LLMs reached high agreement, yet each proxy LLM had low correlation with users' evaluations. These results indicate that proxy LLMs cannot accurately estimate users' wide range of perceptions of utility and privacy in privacy-sensitive scenarios. We discuss the need for more user-centered studies to measure LLMs' ability to help users while preserving privacy, and for improving alignment between LLMs and users in estimating perceived privacy and utility.

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.

Expert domain writing, such as scientific writing, typically demands extensive domain knowledge. Although large language models (LLMs) show promising potential in this task, evaluating the quality of automatically generated scientific writing is a crucial open issue, as it requires knowledge of domain-specific criteria and the ability to discern expert preferences. Conventional automatic evaluation metrics and LLM-as-a-judge systems, primarily designed for mainstream NLP tasks, are insufficient to grasp expert preferences and domain-specific quality standards. To address this gap and support realistic human-AI collaborative writing, we focus on related work generation, one of the most challenging scientific tasks, as an exemplar. We propose GREP, a multi-turn evaluation framework that integrates classical related work evaluation criteria with expert-specific preferences. GREP decomposes the evaluation into smaller fine-grained dimensions. This localized evaluation is further augmented with contrastive examples to provide detailed contextual guidance for the evaluation dimensions. Empirical investigation reveals that GREP is able to assess the quality of related work sections in a much more robust manner compared to standard LLM judges, reflects natural scenarios of scientific writing, and bears a strong correlation with the assessment of human experts. We also observe that generations from state-of-the-art LLMs struggle to satisfy validation constraints of a suitable related work section.

With the surge of large language models (LLMs) and their ability to produce customized output, style-personalized text generation--"write like me"--has become a rapidly growing area of interest. However, style personalization is highly specific, relative to every user, and depends strongly on the pragmatic context, which makes it uniquely challenging. Although prior research has introduced benchmarks and metrics for this area, they tend to be non-standardized and have known limitations (e.g., poor correlation with human subjects). LLMs have been found to not capture author-specific style well, it follows that the metrics themselves must be scrutinized carefully. In this work we critically examine the effectiveness of the most common metrics used in the field, such as BLEU, embeddings, and LLMs-as-judges. We evaluate these metrics using our proposed style discrimination benchmark, which spans eight diverse writing tasks across three evaluation settings: domain discrimination, authorship attribution, and LLM-generated personalized vs non-personalized discrimination. We find strong evidence that employing ensembles of diverse evaluation metrics consistently outperforms single-evaluator methods, and conclude by providing guidance on how to reliably assess style-personalized text generation.

Large language models (LLMs) may support counseling training, yet evidence from Japanese-language interactions and automated quality ratings remains limited. We examined 18 fixed Japanese-language counseling transcripts generated through artificial intelligence (AI)-to-AI interactions under three counselor conditions: GPT-minimal (GPT-4-turbo with a minimal role instruction), GPT-SMDP (GPT-4-turbo with the Structured Multi-step Dialogue Prompt [SMDP]), and Claude-SMDP (Claude-3-Opus with SMDP). Fifteen counseling experts rated transcripts on four adapted global scales from the Motivational Interviewing Treatment Integrity coding manual and an overall-quality item; three newer LLMs independently rated the same transcripts in three iterations. In this fixed stimulus set, SMDP-condition dialogues received higher expert ratings for cultivating change talk, partnership, empathy, and overall quality than GPT-minimal dialogues; the two SMDP counselor models did not differ. LLM ratings were reproducible but generally more lenient than expert-reference ratings, particularly for softening sustain talk and overall quality. Simulated-client naturalness was below the scale midpoint. These findings provide an expert-referenced benchmark for Japanese-language AI counseling simulations and show that reproducible LLM ratings should not be treated as calibrated counseling-quality evidence without expert validation. This study does not test clinical effectiveness or human-client outcomes.

The steady improvements of text-to-image (T2I) generative models lead to slow deprecation of automatic evaluation benchmarks that rely on static datasets, motivating researchers to seek alternative ways to evaluate T2I progress. We present Multimodal Text-to-Image Eval (MT2IE), an evaluation framework in which a single multimodal large language model (MLLM) acts as an evaluator agent, iteratively generating the evaluation prompts and scoring the resulting images. We show that MT2IE's image-text consistency scores have higher correlation with human judgment than metrics previously introduced in the literature. MT2IE generates prompts that are efficient at probing T2I model performance: closely recovering the official T2I model rankings of three structurally distinct benchmarks from just 20 generated evaluation prompts, 28-105x fewer than the benchmarks' own prompt sets. When compared to existing evaluation metrics such as CLIPScore, VIEScore, and VQAScore, MT2IE's T2I model rankings are more faithful and far more consistent across multiple evaluation seeds when using the same number of prompts. MT2IE can also adapt evaluation to the model being tested: rewriting each prompt based on the model's own measured performance to produce a bespoke per-model benchmark that still recovers the official rankings and keeps the evaluated model in an informative scoring range. We hope that these results will encourage the development of dynamic and interactive evaluation frameworks, and mitigate the deprecation of automatic evaluation benchmarks.

Automatic text summarization relies on automatic evaluation to quickly determine the quality of summarization models via automatic metrics and LLM-as-a-Judge models. However, these techniques require meta-evaluation to ensure that they capture human judgments correctly. In this paper, we explore this meta-evaluation beyond English by generating a new multilingual summary meta-evaluation dataset (BASSE), which comprises human judgments on 2,040 abstractive summaries, generated either manually or by five Large Language Models (LLMs) with four different prompts. For each summary, annotators evaluate five criteria on a 5-point Likert scale: coherence, consistency, fluency, relevance, and 5W1H. We then benchmark automatic summarization metrics and LLM-as-a-Judge models. Our results show that currently proprietary judge LLMs have the highest correlation with human judgments, followed by criteria-specific automatic metrics, while open-sourced judge LLMs perform poorly.

Reliable evaluation of large language models (LLMs) is critical as their deployment rapidly expands, particularly in high-stakes domains such as business and finance. The LLM-as-a-Judge framework, which uses prompted LLMs to evaluate response quality, is appealing due to its scalability, low cost, and strong correlations with human stylistic preferences. However, it remains unclear how accurately these methods can assess response quality in domains where correctness matters more than style. To address this gap, we introduce the Business and Finance Fundamentals Benchmark (BFF-Bench), a dataset of 160 challenging questions and long-form responses authored by financial professionals. These experts subsequently evaluated the correctness of 1,200 responses generated by a diverse set of LLMs on both BFF-Bench and a challenging subset of MT-Bench. With this expert-annotated dataset of judgments (VERDICTS), we analyze the agreement between a suite of automated grading methods and human experts. While we observe that LLM Judges are more reliable than other grading methods, our findings reveal a clear pattern in LLM Judge performance: when not provided with a correct reference, judges show high agreement with human experts only on questions the judges were able to correctly answer themselves. We demonstrate that providing the judges with expert-written references largely mitigates this issue, highlighting the limits of using LLM-as-a-Judge without any form of human verification.