One Token to Fool LLM-as-a-Judge

Generative reward models using LLMs for evaluating answer quality are vulnerable to superficial manipulations, prompting the need for improved evaluation methods and a robust new model to enhance reliability.

https://arxiv.org/abs//2507.08794

YouTube: https://www.youtube.com/@ArxivPapers

TikTok: https://www.tiktok.com/@arxiv_papers

Apple Podcasts: https://podcasts.apple.com/us/podcast/arxiv-papers/id1692476016

Spotify: https://podcasters.spotify.com/pod/show/arxiv-papers