Researcher Says TypeSafe's Jev Mirrors His Autorubric LLM Eval Framework From 8 Months Ago

deliprao · x · 2026-09-21

Delip Rao says @typesafeai's new Jev product uses an abstraction strikingly similar to Autorubric, his open-source rubric-based LLM evaluation framework released 8 months ago and to be presented at COLM next month — with Jev's Noul/Score/Choice criterion types mapping 1:1 to Autorubric's. Autorubric (arXiv:2603.00077) unifies rubric-based LLM judging for non-verifiable tasks with bias mitigation, calibration, abstention, and the CHARM-100 benchmark.

Related event: Researcher Alleges TypeSafe's Jev Mirrors His AutoRubric Paper(3 posts)→

Original post →

More from Research

Research channel →