Pith. sign in

REVIEW 41 cited by

Not yet reviewed by Pith; the record is open.

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2310.17631 v2 pith:ZSJAO2BZ submitted 2023-10-26 cs.CL cs.AI

JudgeLM: Fine-tuned Large Language Models are Scalable Judges

classification cs.CL cs.AI
keywords judgelmjudgejudgesagreementbenchmarkbiasllmsmodels
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved
0 comments
read the original abstract

Evaluating Large Language Models (LLMs) in open-ended scenarios is challenging because existing benchmarks and metrics can not measure them comprehensively. To address this problem, we propose to fine-tune LLMs as scalable judges (JudgeLM) to evaluate LLMs efficiently and effectively in open-ended benchmarks. We first propose a comprehensive, large-scale, high-quality dataset containing task seeds, LLMs-generated answers, and GPT-4-generated judgments for fine-tuning high-performance judges, as well as a new benchmark for evaluating the judges. We train JudgeLM at different scales from 7B, 13B, to 33B parameters, and conduct a systematic analysis of its capabilities and behaviors. We then analyze the key biases in fine-tuning LLM as a judge and consider them as position bias, knowledge bias, and format bias. To address these issues, JudgeLM introduces a bag of techniques including swap augmentation, reference support, and reference drop, which clearly enhance the judge's performance. JudgeLM obtains the state-of-the-art judge performance on both the existing PandaLM benchmark and our proposed new benchmark. Our JudgeLM is efficient and the JudgeLM-7B only needs 3 minutes to judge 5K samples with 8 A100 GPUs. JudgeLM obtains high agreement with the teacher judge, achieving an agreement exceeding 90% that even surpasses human-to-human agreement. JudgeLM also demonstrates extended capabilities in being judges of the single answer, multimodal models, multiple answers, multi-turn chat, etc. Code is available at https://github.com/baaivision/JudgeLM.

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 41 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. Self-GC: Self-Governing Context for Long-Horizon LLM Agents

    cs.AI 2026-07 unverdicted novelty 7.0

    Self-GC governs agent context as indexed objects with planner-proposed actions, achieving 84.85% no-impact on future continuations on a hard set versus 54-70% for baselines.

  2. The Hidden Bias of Process Reward Models:PRISM for Rewarding the Right Reasoning

    cs.LG 2026-06 unverdicted novelty 7.0

    PRISM is a contrastive, policy-aware training framework for process reward models that reduces false positives by 22% on PRMBench and boosts downstream accuracy up to 33% in Best-of-N selection by learning reliable re...

  3. AndroidDaily: A Verifiable Benchmark for Mobile GUI Agents on Real-World Closed-Source Applications

    cs.CV 2026-05 unverdicted novelty 7.0

    AndroidDaily supplies 350 verifiable tasks on 94 closed-source Android apps evaluated by GRADE (87.37% human agreement), with the strongest model achieving 62% success.

  4. Prism-Reranker: Beyond Relevance Scoring -- Jointly Producing Contributions and Evidence for Agentic Retrieval

    cs.IR 2026-04 accept novelty 7.0

    Prism-Reranker models output relevance, contribution statements, and evidence passages to support agentic retrieval beyond scalar scoring.

  5. Unveiling Fine-Grained Visual Traces: Evaluating Multimodal Interleaved Reasoning Chains in Multimodal STEM Tasks

    cs.CV 2026-04 unverdicted novelty 7.0

    StepSTEM benchmark and dynamic-programming step alignment show top MLLMs achieve only 38.29% accuracy on graduate STEM tasks requiring interleaved cross-modal reasoning.

  6. Culture-Aware Humorous Captioning: Multimodal Humor Generation across Cultural Contexts

    cs.CL 2026-04 unverdicted novelty 7.0

    Introduces culture-aware humorous captioning task and staged alignment framework that improves contextual fit and balances image relevance with humor in multimodal LLMs.

  7. Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models

    cs.CL 2024-04 conditional novelty 7.0

    A panel of smaller diverse LLMs outperforms a single large model as an evaluator of generations, showing less intra-model bias and over 7x lower cost.

  8. Towards Spec Learning: Inference-Time Alignment from Preference Pairs

    cs.CL 2026-06 unverdicted novelty 6.0

    Proposes compiling preference pairs into readable natural-language specifications for inference-time LLM alignment, claiming outperformance over DPO on dense-preference domains.

  9. IPO Finance Agent: Benchmark of LLM Financial Analysts Beyond Finance Agent v2, with Automated Rubric Generation, on the SpaceX (SPCX) IPO

    cs.AI 2026-06 unverdicted novelty 6.0

    IPO Finance Agent creates a new LLM benchmark for IPO due diligence on SpaceX S-1 with contextual retrieval and automated rubrics, where Qwen 3.7 Max scores 79.4% at $0.30/query, outperforming Finance Agent v2 on accu...

  10. IPO Finance Agent: Benchmark of LLM Financial Analysts Beyond Finance Agent v2, with Automated Rubric Generation, on the SpaceX (SPCX) IPO

    cs.AI 2026-06 unverdicted novelty 6.0

    IPO Finance Agent benchmarks LLMs on SpaceX S-1 questions with contextual retrieval and auto-generated rubrics, reporting up to 79.8% accuracy and better cost-efficiency than prior Finance Agent v2 entries.

  11. Automated Creativity Evaluation of Language Models Across Open-Ended Tasks

    cs.CL 2026-06 unverdicted novelty 6.0

    Authors propose a new framework for automated LLM creativity evaluation that separates measurement from the task, using semantic entropy and multi-agent judges, validated on problem-solving, research ideation, and cre...

  12. Benchmark Everything Everywhere All at Once

    cs.AI 2026-06 unverdicted novelty 6.0

    Benchmark Agent is an autonomous agentic system that constructs benchmarks for LLMs and MLLMs via query analysis, subtask design, annotation and quality control, yielding 15 benchmarks with minimal human input.

  13. A Finite-Calibration Regime Map for LLM Judge Panels

    cs.CL 2026-05 unverdicted novelty 6.0

    The paper introduces a finite-calibration regime map and Finite-Calibration Panel Selection selector, finding scalar aggregation wins on most real benchmark-budget combinations while joint tables help when interaction...

  14. "I Strongly Suspect This Website Is a Scam": Benchmarking PII Leakage and Detection without Defense in Autonomous Web Agents

    cs.CR 2026-05 unverdicted novelty 6.0

    New benchmark Scammer4U finds 54-93% critical PII leakage from frontier web agents on scam sites versus 0% on benign twins, plus a 30-point gap between verbalized suspicion and actual submission.

  15. DeepSurvey: Enhancing Analytical Depth and Citation Reliability in Automated Survey Generation

    cs.AI 2026-05 unverdicted novelty 6.0

    DeepSurvey introduces an agentic system for automated survey generation that improves depth through full-text keynotes, cross-paper clustering, and code analysis, while boosting citation reliability via graph expansio...

  16. Agreement Metrics for LLM-as-Judge Evaluation: What to Report and Why

    cs.CL 2026-05 conditional novelty 6.0

    For binary LLM judge validation, Pearson's r, Spearman's ρ, Kendall's τ_b, phi, and Matthews correlation all equal a single number on non-degenerate data, Cohen's κ supplies the extra signal on label-rate drift, and a...

  17. Unveiling Fine-Grained Visual Traces: Evaluating Multimodal Interleaved Reasoning Chains in Multimodal STEM Tasks

    cs.CV 2026-04 unverdicted novelty 6.0

    StepSTEM benchmark and step-level DP evaluation show top MLLMs achieve only 38.29% accuracy on fine-grained multimodal STEM reasoning, relying primarily on textual cues.

  18. HoWToBench: Holistic Evaluation for LLM's Capability in Human-level Writing using Tree of Writing

    cs.CL 2026-04 unverdicted novelty 6.0

    Tree-of-Writing achieves 0.93 Pearson correlation with human judgments by using a tree-structured workflow to aggregate sub-feature scores, outperforming standard LLM-as-a-judge and overlap metrics on the new HowToBench.

  19. AdaRubric: Task-Adaptive Rubrics for Reliable LLM Agent Evaluation and Reward Learning

    cs.AI 2026-03 conditional novelty 6.0

    AdaRubric adaptively generates task-specific rubrics via LLM, scores agent trajectories with per-dimension confidence weighting, and produces filtered DPO pairs that raise human correlation to Pearson r=0.79 and downs...

  20. ACE-Bench: A Lightweight Benchmark for Evaluating Azure SDK Usage Correctness

    cs.DC 2026-02 unverdicted novelty 6.0

    ACE-Bench is an execution-free benchmark that scores LLM coding agents on correct Azure SDK usage via deterministic regex checks and reference-based LLM judges derived from official documentation.

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

    cs.CL 2026-07 conditional novelty 5.0

    Judge upgrades are not interchangeable: only Qwen3 1.7B→4B yields robust adjacent gains, MiniMax adjacent releases do not, and stronger judges reduce but do not remove bias or correlated jury errors.

  22. Towards Spec Learning: Inference-Time Alignment from Preference Pairs

    cs.CL 2026-06 unverdicted novelty 5.0

    Spec learning compiles brief instructions and preference pairs into readable natural-language specifications that condition LLMs at inference time and can outperform DPO on domains with dense preference signals.

  23. AURA: Adaptive Uncertainty-aware Refinement for LLM-as-a-Judge Auditing

    stat.ML 2026-06 unverdicted novelty 5.0

    AURA is an adaptive uncertainty-aware refinement method for auditing LLM-as-a-judge pairwise decisions that learns human-consistency signals through selective human verification on uncertain cases.

  24. Quantifying and Auditing LLM Evaluation via Positive--Unlabeled Learning

    stat.ML 2026-06 unverdicted novelty 5.0

    A positive-unlabeled learning approach using partial optimal transport is introduced to audit and correct biases in LLM-as-a-judge systems by aligning limited human positives with unlabeled outputs in embedding space.

  25. Beyond Scalar Scores: Exploring LLM-based Metrics for Clinical Significance Evaluation in Radiology Reports

    cs.CL 2026-06 unverdicted novelty 5.0

    Lightweight metrics trained on Qwen3-8B and MedGemma-4B using synthetic pairs outperform larger medical LLMs at distinguishing clinical significance in radiology reports while balancing discrimination and robustness.

  26. Prompt Perturbation for Reliable LLM Evaluation over Comparison Graphs

    cs.CL 2026-06 unverdicted novelty 5.0

    A prompt perturbation approach builds comparison graphs from LLM judgments, filters inconsistent cycles or ties, and aggregates more reliable rankings.

  27. MemSlides: A Hierarchical Memory Driven Agent Framework for Personalized Slide Generation with Multi-turn Local Revision

    cs.CL 2026-06 unverdicted novelty 5.0

    MemSlides introduces a three-part memory hierarchy (user profile, working, tool) with scoped local revision for multi-turn personalized slide generation.

  28. Supporting System Testing with a Multi-Agent LLM-based Framework for Knowledge Graph Extraction: A Case Study with Ethernet Switch Systems

    cs.SE 2026-05 conditional novelty 5.0

    A multi-agent LLM-based framework extracts knowledge graphs from 50 real Ethernet switch manuals with 0.97-0.99 correctness to enable downstream test case specification generation.

  29. LLM-as-a-Judge for Human-AI Co-Creation: A Reliability-Aware Evaluation Framework for Coding

    cs.SE 2026-04 unverdicted novelty 5.0

    LLM judges for human-AI coding co-creation show moderate performance (ROC-AUC 0.59) and low agreement, with co-creation success concentrating early in interactions.

  30. How Long Reasoning Chains Influence LLMs' Judgment of Answer Factuality

    cs.CL 2026-04 unverdicted novelty 5.0

    Weak LLM judges accept wrong answers more often when shown fluent reasoning chains, while strong judges use them partially but still get misled by high-quality-looking but flawed reasoning.

  31. MedConclusion: A Benchmark for Biomedical Conclusion Generation from Structured Abstracts

    cs.CL 2026-04 unverdicted novelty 5.0

    MedConclusion is a 5.7M-instance benchmark dataset for generating biomedical conclusions from structured PubMed abstracts, with LLM evaluations showing conclusion writing differs from summarization and that judge choi...

  32. Breakdowns in Conversational AI: Interactional Failures in Emotionally and Ethically Sensitive Contexts

    cs.CL 2026-04 unverdicted novelty 5.0

    Mainstream conversational models show escalating affective misalignments and ethical guidance failures during staged emotional trajectories, organized into a taxonomy of interactional breakdowns.

  33. SenseJudge: Human-Centric Preference-Driven Judgment Framework

    cs.CL 2026-06 unverdicted novelty 4.0

    SenseJudge proposes a human-preference-driven LLM judgment framework and SenseBench benchmark that claims to outperform prior methods in personalized judging and produce human-aligned model rankings.

  34. OmniVerifier-M1: Multimodal Meta-Verifier with Explicit Structured Recalibration

    cs.CL 2026-05 unverdicted novelty 4.0

    OmniVerifier-M1 is a generalist visual verifier using symbolic outputs for meta-verification and decoupled RL to outperform joint optimization for robust verification and agentic self-correction.

  35. Refining and Reusing Annotation Guidelines for LLM Annotation

    cs.CL 2026-05 conditional novelty 4.0

    An iterative moderation framework refines and reuses annotation guidelines to improve LLM annotation accuracy on biomedical NER tasks across GPT, Gemini, and DeepSeek models.

  36. FinRAG-12B: A Production-Validated Recipe for Grounded Question Answering in Banking

    cs.AI 2026-05 unverdicted novelty 4.0

    FinRAG-12B is a production-deployed 12B model for banking that grounds answers with citations, refuses unanswerable queries at a calibrated 12% rate, outperforms GPT-4.1 on grounding, and improves query resolution by ...

  37. LLM-as-Judge for Semantic Judging of Powerline Segmentation in UAV Inspection

    cs.AI 2026-04 unverdicted novelty 4.0

    An LLM produces consistent categorical judgments and appropriate confidence declines when evaluating powerline segmentation quality under controlled visual degradations, suggesting it can serve as a reliable watchdog.

  38. Synthetic Eggs in Many Baskets: The Impact of Synthetic Data Diversity on LLM Fine-Tuning

    cs.CL 2025-11 unverdicted novelty 4.0

    Fine-tuning LLMs on multi-source synthetic data mitigates distribution collapse and self-preference bias while increasing output quality relative to single-source or human-only fine-tuning.

  39. A Survey on LLM-as-a-Judge

    cs.CL 2024-11 unverdicted novelty 4.0

    A survey on LLM-as-a-Judge that reviews reliability strategies, proposes evaluation methods, and introduces a novel benchmark for assessing such systems.

  40. Evaluating RAG Metrics in Applied Contexts: An Experiment, Its Findings and Its Limitations

    cs.CL 2026-07 conditional novelty 3.0

    Correlation analysis of four RAG metric libraries against human evaluations on 96 questions shows RAGChecker metrics correlate strongly with human scores, while generation-as-overall metrics correlate weakly, but sing...

  41. A Primer in Post-Training Reasoning Data: What We Know About How It Works

    cs.CL 2026-06 unverdicted novelty 2.0

    A literature synthesis that organizes post-training reasoning data research around data objects, usefulness factors, construction methods, and scaling behaviors to create an attribution framework.