mlsanityData to 14 Oct 2026

MCJudgeBench: A Benchmark for Constraint-Level Judge Evaluation in Multi-Constraint Instruction Following

arXiv 2605.03858 ↗First posted 5 May 2026cs.CL cs.AIcited by 0

Not on the map. Why?

Abstract (arXiv)

Multi-constraint instruction following requires verifying whether a response satisfies multiple individual requirements, yet LLM judges are often assessed only through overall-response judgments. We introduce MCJudgeBench, a benchmark for constraint-level judge evaluation in multi-constraint instruction following. Each instance includes an instruction, a candidate response, an explicit constraint list, per-constraint gold labels in {yes, partial, no}, and controlled response-side perturbations. The evaluation protocol further includes evaluation prompt variants to test judge stability. We evaluate proprietary and open-source LLM judges using both correctness and inconsistency metrics, distinguishing intrinsic inconsistency under stochastic decoding from procedural inconsistency under prompt and response perturbations. Our results show that judge reliability has multiple dimensions: strong overall performance does not guarantee equally reliable detection across label categories, especially for rarer partial and no cases. Judges with higher correctness do not always have lower inconsistency. Evaluation with reasoning improves correctness but does not uniformly improve stability. These findings motivate evaluating LLM judges at the constraint level to study these failure modes.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Builds on 19 of its 25 references that are in the map. First posted on arXiv 5 May 2026.

Builds on

19 of 25 references.

References · 19

  1. 2024The Llama 3 Herd of Modelscited by 14,715
  2. 2023Judging LLM-as-a-Judge with MT-Bench and Chatbot Arenacited by 8,626
  3. 2020Beyond Accuracy: Behavioral Testing of NLP models with CheckListcited by 1,006
  4. 2023Prometheus: Inducing Fine-grained Evaluation Capability in Language Modelscited by 575
  5. 2024Justice or Prejudice? Quantifying Biases in LLM-as-a-Judgecited by 368
  6. 2024JudgeBench: A Benchmark for Evaluating LLM-based Judgescited by 318
  7. 2023Evaluating Large Language Models at Evaluating Instruction Followingcited by 302
  8. 2024Self-Preference Bias in LLM-as-a-Judgecited by 258
  9. 2024Large Language Models are Inconsistent and Biased Evaluatorscited by 153
  10. 2024Benchmarking Complex Instruction-Following with Multiple Constraints Compositioncited by 152
  11. 2024InFoBench: Evaluating Instruction Following Ability in Large Language Modelscited by 135
  12. 2024Mind Your Step (by Step): Chain-of-Thought can Reduce Performance on Tasks where Thinking Makes Humans Worsecited by 109
7 more
  1. 2024CFBench: A Comprehensive Constraints-Following Benchmark for LLMscited by 67
  2. 2021Perturbation CheckLists for Evaluating NLG Evaluation Metricscited by 50
  3. 2025LLMs Cannot Reliably Judge (Yet?): A Comprehensive Assessment on the Robustness of LLM-as-a-Judgecited by 37
  4. 2024Evaluating the Consistency of LLM Evaluatorscited by 24
  5. 2024ReIFE: Re-evaluating Instruction-Following Evaluationcited by 22
  6. 2024GroUSE: A Benchmark to Evaluate Evaluators in Grounded Question Answeringcited by 5
  7. 2026FairJudge: An Adaptive, Debiased, and Consistent LLM-as-a-Judgecited by 5

Built on by

Not cited yet.