REVIEW 19 cited by
Judging the Judges: Evaluating Alignment and Vulnerabilities in LLMs-as-Judges
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
Judging the Judges: Evaluating Alignment and Vulnerabilities in LLMs-as-Judges
read the original abstract
Offering a promising solution to the scalability challenges associated with human evaluation, the LLM-as-a-judge paradigm is rapidly gaining traction as an approach to evaluating large language models (LLMs). However, there are still many open questions about the strengths and weaknesses of this paradigm, and what potential biases it may hold. In this paper, we present a comprehensive study of the performance of various LLMs acting as judges, focusing on a clean scenario in which inter-human agreement is high. Investigating thirteen judge models of different model sizes and families, judging answers of nine different 'examtaker models' - both base and instruction-tuned - we find that only the best (and largest) models achieve reasonable alignment with humans. However, they are still quite far behind inter-human agreement and their assigned scores may still differ with up to 5 points from human-assigned scores. In terms of their ranking of the nine exam-taker models, instead, also smaller models and even the lexical metric contains may provide a reasonable signal. Through error analysis and other studies, we identify vulnerabilities in judge models, such as their sensitivity to prompt complexity and length, and a tendency toward leniency. The fact that even the best judges differ from humans in this comparatively simple setup suggest that caution may be wise when using judges in more complex setups. Lastly, our research rediscovers the importance of using alignment metrics beyond simple percent alignment, showing that judges with high percent agreement can still assign vastly different scores.
Forward citations
Cited by 19 Pith papers
-
OR-Space: A Full-Lifecycle Workspace Benchmark for Industrial Optimization Agents
OR-Space is a benchmark for LLM agents performing full-lifecycle optimization tasks across Build, Revise, and Explain modes in executable multi-artifact workspaces.
-
Evaluating Tool-Using Language Agents: Judge Reliability, Propagation Cascades, and Runtime Mitigation in AgentProp-Bench
AgentProp-Bench shows substring judging agrees with humans at kappa=0.049, LLM ensemble at 0.432, bad-parameter injection propagates with ~0.62 probability, rejection and recovery are independent, and a runtime fix cu...
-
Code Review Agent Benchmark
c-CRAB benchmark shows state-of-the-art code review agents solve only around 40% of tasks derived from human reviews, suggesting potential for human-AI collaboration.
-
NUTSHELL: A Dataset for Abstract Generation from Scientific Talks
NUTSHELL is a new open dataset of ACL talks paired with abstracts, accompanied by baselines that demonstrate training benefits for speech-to-abstract generation while highlighting remaining challenges.
-
POIROT: Interrogating Agents for Failure Detection in Multi-Agent Systems
POIROT protocol repurposes agents in LLM multi-agent systems as an internal diagnostic layer for failure detection, outperforming single-LLM evaluators with gains that increase with complexity, agent count, and fault types.
-
Navigating the Sea of LLM Evaluation: Investigating Bias in Toxicity Benchmarks
Toxicity benchmarks for LLMs produce inconsistent results when task type, input domain, or model changes, revealing intrinsic evaluation biases.
-
Effective Performance Measurement: Challenges and Opportunities in KPI Extraction from Earnings Calls
Encoder models trained on SEC filings struggle with earnings calls due to domain shift, while LLMs enable open-ended KPI extraction with 79.7% human-verified precision on newly introduced benchmarks.
-
LH-Bench: Skill-Grounded Evaluation of Long-Horizon Agents on Subjective Enterprise Tasks
Expert-authored dual-use skills make LLM judges more reliable on subjective long-horizon enterprise agent tasks than LLM-authored rubrics, with human preferences confirming the main ranking boundary.
-
Conversation for Non-verifiable Learning: Self-Evolving LLMs through Meta-Evaluation
CoNL lets LLMs self-improve on non-verifiable tasks by rewarding critiques that produce better solutions in multi-agent conversations, jointly optimizing generation and judging without external feedback.
-
Reinforcement Learning with Verifiable yet Noisy Rewards under Imperfect Verifiers
Derives backward and forward corrections for asymmetric verifier noise that improve RLVR performance on math reasoning tasks.
-
AURA: Adaptive Uncertainty-aware Refinement for LLM-as-a-Judge Auditing
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.
-
Mixed response geometry and critical crossover in the Ising model
Thermodynamic curvature on the (β, h) manifold in the Ising model produces a ridge that geometrically identifies the Widom line as the locus of maximal response extending from the critical point.
-
Mixed response geometry and critical crossover in the Ising model
In the 2D Ising model, the mixed response Ω_βh = −N cov(m,e) forms a localized ridge from criticality into finite-field crossover and collapses onto a susceptibility-constrained curve in normalized coordinates.
-
WildFeedback: Aligning LLMs With In-situ User Interactions And Feedback
WildFeedback extracts preference pairs from in-situ user feedback in LLM conversations to fine-tune models for better alignment with real user preferences.
-
Multi-Dimensional Evaluation of Sustainable City Trips with LLM-as-a-Judge and Human-in-the-Loop
LLMs judging sustainable city-trip recommendations show model-specific biases and high variance across four dimensions even when overall rankings agree, with a three-phase human calibration process clarifying some rea...
-
A Survey on LLM-as-a-Judge
A survey on LLM-as-a-Judge that reviews reliability strategies, proposes evaluation methods, and introduces a novel benchmark for assessing such systems.
-
From Cool Demos to Production-Ready FMware: Core Challenges and a Technology Roadmap
A semi-structured thematic synthesis identifies core challenges in FM selection, alignment, prompting, orchestration, testing, deployment, and cross-cutting concerns like observability for production-ready FMware.
-
Multi-Stage Retrieval for Operational Technology Cybersecurity Compliance Using Large Language Models: A Railway Casestudy
A parallel compliance architecture using multi-stage LLM retrieval improves correctness and reasoning quality over a baseline for OT cybersecurity compliance queries in a railway case study.
-
LLMs-as-Judges: A Comprehensive Survey on LLM-based Evaluation Methods
A survey that organizes LLMs-as-judges research into functionality, methodology, applications, meta-evaluation, and limitations.
discussion (0)
Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.