Text-to-Image Models Leave Identifiable Signatures: Implications for Leaderboard... | Hacker News

Hacker Newsnew | past | comments | ask | show | jobs | submit

		nycdatasci 5 days ago \| parent \| context \| favorite \| on: GPT Image 1.5 Text-to-Image Models Leave Identifiable Signatures: Implications for Leaderboard Security https://arxiv.org/pdf/2510.06525

encroach 5 days ago [–]

This is true, however LMArena does employ some methods to mitigate attempts to manipulate the leaderboard, see https://openreview.net/forum?id=zf9zwCRKyP

They also control for style https://news.lmarena.ai/sentiment-control/

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact