The paper introduces an end-to-end evaluation pipeline for large language model (LLM) software systems that integrates checklist creation, learned aggregation of checklist responses, and additional features such as self‑consistency, explanations, and prediction uncertainty. This pipeline aims to enhance agreement among LLM judges and improve alignment with human judgments, addressing practical reliability concerns that previous work has only partially covered. Empirical results demonstrate the effectiveness of the proposed framework.
arXiv:2609.24516v1 Announce Type: new Abstract: In recent years, large language models (LLMs) have emerged as a popular alternative for evaluation. Often referred to as LLMs as judges (LLJs), these s...
arXiv:2608. 08700v1 Announce Type: new Abstract: Reliable evaluation of tool routing is critical as Large Language Models increasingly operate as autonomous agents.
ProofJudge is an agentic large language model that evaluates the quality of formal proofs in Lean 4 beyond mere correctness. It scores proofs on five dimensions—library leverage, automation fit, structural clarity, statement quality, and Mathlib conventions—using tool access to the relevant repository commit. The system was tested on 218 Mathlib declarations, achieving alignment with human reviewers between 63.5% and 80.8% and releasing its harness, dataset, and traces for open research.
The paper argues that answer accuracy alone is insufficient for evaluating large language model (LLM) data agents, especially in structured-data tasks where a correct answer can be produced by an invalid trace. It introduces Trace Integrity as a reliability criterion that ensures the computation behind an answer is explicit, executable, schema-valid, operator-faithful, replayable, answer-consistent, and auditable. The authors operationalize this concept with execution contracts and present the CAIT (Correct Answer / Invalid Trace) Rate to quantify how often answer-only evaluations mistakenly reward unsupported outputs, demonstrating that accuracy, trace validity, and silent-failure risk are distinct signals.
The paper examines LLM-as-a-Judge systems used to assess AI-generated text, questioning the assumption that judgments are derived from reasoning over responses and rubrics. It finds that classifiers trained solely on rubric text can predict judge outputs, indicating that rubrics contain recoverable evaluative signals independent of the responses. Counterfactual experiments show judges often fail to adjust decisions when either the response or rubric criterion is reversed, raising doubts about the reliability of rubric-based LLM evaluation.
arXiv:2604. 22891v4 Announce Type: replace-cross Abstract: LLM-as-a-Judge has become a dominant approach in automated evaluation systems, playing critical roles in model alignment, leaderboard construction, quality control, and so on.
The paper introduces LongJudgeBench, a benchmark designed to evaluate large language models (LLMs) acting as judges for long-form text generation. It highlights that long-form evaluation requires complex, document-level assessments beyond simple length, such as organization, coverage, depth, consistency, and scenario-specific quality. Experiments show a significant reliability gap among current LLM judges, indicating instability across scenarios and limited effectiveness of rubrics or references.
arXiv:2607. 20455v1 Announce Type: cross Abstract: Human-annotated data remains fundamental to training frontier Large Language Models (LLMs).
Formal proofs in Lean 4 that pass the kernel's type checker can nonetheless vary widely in quality. We introduce ProofJudge, an agentic LLM-as-judge system that scores formal proof quality along five...
Answer accuracy is an insufficient reliability signal for LLM data agents. In structured-data tasks, a benchmark-correct answer can be produced by an invalid trace. This paper introduces Trace Integri...
arXiv:2606. 16000v1 Announce Type: cross Abstract: We introduce GRACE-DS, a Guarded Reward-guided Agent Correction Environment in Data Science for pre-deployment evaluation of LLM-powered AutoML agents.
arXiv:2608. 04439v1 Announce Type: cross Abstract: Large language models (LLMs) have made notable progress in code generation, but they still struggle on challenging tasks that require sophisticated algorithms or complex implementations.