Reviewer calls for AI model evals to focus on daily cognitive work, not flashy demos
vivekhaldar · x · 2026-11-12
vivekhaldar publicly asked Matthew Berman, Peter Yang, and LMArena to shift model reviews away from fantastical 3D civilization walkthroughs and toward the daily cognitive work people actually use models for, arguing current eval showcases have drifted from real-world usage.
Related event: Critics Call for AI Benchmarks to Focus on Everyday Cognitive Tasks(2 posts)→