Metraipolis › archive › dc_4d218ff63fb5

All Verdicts are Not Equal: Rethinking LLM Judge Reliability — arXiv:2610.12083

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.12083
License (as archived)
CC0-1.0
Archived
2026-10-09T06:03:28.843Z
SHA-256 of the archived text
0f0a2d03c8519405a5140a4d1e9bc27b50a1770fbf1f24154234006fedd4b2b7
Stable URL of this page
https://metraipolis.com/documento/dc_4d218ff63fb5

How to cite

All Verdicts are Not Equal: Rethinking LLM Judge Reliability — arXiv:2610.12083 — arXiv (en), https://arxiv.org/abs/2610.12083 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_4d218ff63fb5 (Metraipolis archive dc_4d218ff63fb5, sha256 0f0a2d03c8519405a5140a4d1e9bc27b50a1770fbf1f24154234006fedd4b2b7)
BibTeX
@misc{dc4d218ff63fb5,
  title = {All Verdicts are Not Equal: Rethinking LLM Judge Reliability — arXiv:2610.12083},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.12083},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_4d218ff63fb5; sha256 0f0a2d03c8519405a5140a4d1e9bc27b50a1770fbf1f24154234006fedd4b2b7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.12083. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

LLM-as-a-Judge is the standard paradigm for NLP evaluation, yet its systemic reliability remains poorly understood despite being widely treated as a deterministic ground truth. We present a comprehensive reliability audit, stresstesting six frontier models across four benchmarks, five prompt formats, two presentation orders, three sampling temperatures, and ten repetitions per condition. Our empirical analysis reveals severe vulnerabilities: verdicts change across identical replications at temperature zero, position-order swaps flip the majority of verdicts on challenging tasks, and the most deterministic judge achieves perfect consistency by trivially repeating incorrect verdicts, agreeing with ground truth only 51% of the time. To formalize these multi-faceted failure modes, we introduce the trustworthy verdict rate (T ), a unified metric capturing the joint probability that an evaluation is reproducible, order-invariant, and accurate. UsingT , we derive a theoretical upper bound on accuracy imposed by position bias and show that reliability is item-specific rather than modellevel. Finally, we demonstrate that shifting from pairwise win-rate to holistic rubric scoring improves trustworthiness more than any single-format prompting intervention, offering an actionable framework for robust NLP evaluation.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.12083 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-09T05:47:02.011Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).