Skip to content

Research

DeepThink research data open for download

Rogue published the aggregates behind the DeepThink Preview charts as CSV, JSON and BibTeX, for anyone to check and cite.

Rogue DeepThink Preview: 75.22% on DeepSWE v1.1

Rogue opened the aggregate data behind the Rogue DeepThink Preview research: 17 effort and cost points across four benchmarks, including the GLM 5.3 baseline, and 9 patch-size means on 77 matched DeepSWE v1.1 tasks.

The package ships as CSV and JSON with a BibTeX citation, next to the definition of each benchmark. It is the same data that draws the charts in the technical article.

More news

All news