Restricted / Non-commercial license tags with links; fev-bench pairwise heatmaps

#1
by jborn50 - opened
  • Reads commercial_use as yes / no / restricted (old true/false files still work); Timestral-209M shows a Restricted tag and TimesFM-3.0 a Non-commercial tag, each linking to its license. The filter reads "Include non-commercial and restricted models".
  • Adds a Pairwise comparison section: fev-bench's win-rate and skill-score heatmaps, 95% CIs on hover, following the board and metric selectors (Vega 5.30.0 / Vega-Lite 5.21.0 / vega-embed 6.26.0).

Reads pairwise.csv and license_url from the dataset: merge together with https://huggingface.co/datasets/carbon-arc/RealCast-bench/discussions/1

jborn50 changed pull request status to merged

Sign up or log in to comment