Metraipolis › archive › dc_3c1d2327bfca
From Judgment Quality to Downstream Utility: Rethinking LLM-as-a-Judge for Open-Ended Tasks — arXiv:2609.37145
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.37145
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-07T16:23:32.570Z
- SHA-256 of the archived text
4f1be77ffcd21c012d5e09df4b99f24ffee061e28f33696b7d5a9ce4f8c0db7f- Stable URL of this page
https://metraipolis.com/documento/dc_3c1d2327bfca
How to cite
From Judgment Quality to Downstream Utility: Rethinking LLM-as-a-Judge for Open-Ended Tasks — arXiv:2609.37145 — arXiv (en), https://arxiv.org/abs/2609.37145 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_3c1d2327bfca (Metraipolis archive dc_3c1d2327bfca, sha256 4f1be77ffcd21c012d5e09df4b99f24ffee061e28f33696b7d5a9ce4f8c0db7f)
BibTeX
@misc{dc3c1d2327bfca,
title = {From Judgment Quality to Downstream Utility: Rethinking LLM-as-a-Judge for Open-Ended Tasks — arXiv:2609.37145},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.37145},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_3c1d2327bfca; sha256 4f1be77ffcd21c012d5e09df4b99f24ffee061e28f33696b7d5a9ce4f8c0db7f},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.37145. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
LLM-as-a-Judge is increasingly used to evaluate policy responses on open-ended tasks that lack ground-truth answers. Existing work often directly converts the resulting judgments into reward signals for policy training, paying limited attention to intrinsic judgment quality and largely restricting the use of Judges to training-time supervision. We systematically investigate judgment quality and downstream utility by examining both how judgments are elicited and how they are used. For judgment elicitation, we vary the Judge protocol along three dimensions: verdict granularity, critique usage, and evaluation batching. For judgment usage, beyond policy training, we extend Judge to test-time inference through Best-of-N selection, Judge-guided revision, and beam search. We find that, (i) Surprisingly, judgment quality and downstream utility do not always align. (ii) Judge protocol design substantially affects both intrinsic judgment quality and downstream utility. (iii) Judge guidance effectively converts test-time compute into performance gains, with benefits varying across inference strategies. Our results call for a multifaceted evaluation of LLM Judges on open-ended tasks, encompassing intrinsic judgment quality, and downstream utility. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI). Categorie arXiv: Artificial Intelligence (cs.AI). Pagina del lavoro: https://arxiv.org/abs/2609.37145 — 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 2026-10-07T15:54:53.593Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).