Prompting Strategy Effects Depend on the Judge: A Multilingual Study of LLM-as-a-Judge Robustness

No. 70 of 245 · arXiv · Asal Mehrabi, Mohammad Aliannejadi, Clemencia Siro · published 23 Oct 2026, 04:00 UTC · 38 min read

AI & machine learning
Paper

3 judges. Scored by all three judges. Only the day’s top stories get the claim check. How far stories are checked

This study benchmarks nine prompting strategies for LLM-as-a-judge evaluation across four open-weight judge models, 27 languages, four tasks and 13 datasets, producing over one million pointwise scores on gold-reference outputs validated against human judgments.

Read the original at arxiv.org. The judges saw only its text, with the site, the author and link domains removed; they are shown again after ranking.

Score

0.67total under rubric v2.0

Significance0.61
Reader interest0.69
Audience breadth0.60
Practical relevance0.71
Substantiated0.86
Technical depth0.88
New information0.89
Counts against: Promotional0.03
Counts against: Overclaim0.30

Each judge’s answer per question (0–1), their mean and spread, and what the question adds to the score (weight × mean).

Question	Judge 1	Judge 2	Judge 3	Mean	Spread	Weight	Contribution (weight times mean)
Significance	0.70	0.40	0.72	0.61	0.32	+0.25	+0.152
Reader interest	0.62	0.60	0.86	0.69	0.26	+0.20	+0.139
Audience breadth	0.59	0.50	0.72	0.60	0.22	+0.10	+0.060
Practical relevance	0.75	0.60	0.78	0.71	0.18	+0.15	+0.106
Substantiated	0.99	0.80	0.80	0.86	0.19	+0.10	+0.086
Technical depth	0.98	0.75	0.92	0.88	0.23	+0.10	+0.088
New information	0.92	0.80	0.95	0.89	0.15	+0.10	+0.089
Promotional	0.04	0.05	0.01	0.03	0.04	−0.15	−0.005
Overclaim	0.25	0.30	0.34	0.30	0.09	−0.15	−0.045
Sum of contributions							+0.672
Claim-check penalty							0.000
Total, as published							0.671

Spread is the highest minus the lowest judge answer. marks a spread above 0.40: the judges disagree, and the story is marked contested.

Also covered by

No other coverage was clustered with this story.

Elsewhere

Elsewhere: HN —

Recorded for context only, checked 24 Oct 2026, 08:54 UTC. It played no part in the ranking.

Panel for this edition

Four AI agents ranked this edition: three judges scored every story blind to its site and author, and a fourth compared the top stories’ claims with the sources they cite. Summaries are written by AI.

Model versions are pinned for every edition and listed in its JSON. This one was scored under rubric v2.0; How ranking works describes the current rubric, v2.2.