Open Models Almanac
Methodsevaluationjudge

single source · evaluation

Multimodal LLM instruction-adherence judging

A multimodal LLM assesses sampled video frames for instruction adherence, including for filtering generated videos or trajectories.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See judge for the methods filed beside it.

Filed under evaluation › judge