trends.barnas.net

method · first seen 14 Oct 2026

evaluation-driven benchmark evolution

Continually generating new benchmark tasks from diagnostic feedback on model failures.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

AI agent evaluation