Atlas

Benchmarks

← All benchmarks

ZeroBench Subquestions (dots.vlm1 report)

Multimodal

ZeroBench-sub score reported in the original dots.vlm1 instruction-model card. The card reports a value of 26.65 but does not identify the dataset revision, aggregation method, decoding settings or tool condition for this row. Preserve the original reported metric as a diagnostic; do not conflate it with the versioned official grouped subquestion mean or infer its protocol from comparison columns.

Top models (higher is better)

ModelScore
dots.vlm1 Instruct26.6
Loading Atlas data…