The paper highlights the lack of reliable benchmarks for large language models, proposing "platinum benchmarks" to minimize label errors and revealing persistent model failures in simple tasks.

https://arxiv.org/abs//2502.03461

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