-
docs: link LDR benchmarks dataset for local model selection (#3780)
frostbyte_neo 发布于
2026-05-01 22:30:29 +00:00 - docs: link LDR benchmarks dataset for local model selection
Add prominent links to the community-maintained ldr-benchmarks dataset on
Hugging Face so users can compare local models before downloading them:- README: dedicated "Picking a model?" subsection in Performance, plus a
callout in the Local Models section where users decide what to run. - Research (start) page: small help line under the model dropdown linking
to the dataset, framed around picking a local model.
- docs: point SimpleQA accuracy badge to HuggingFace dataset
The 95% badge previously linked to the ldr-benchmarks GitHub repo
(submissions side). Users clicking 'SimpleQA ~95%' expect to see the
actual leaderboard — that's what the HuggingFace dataset shows, with
sortable CSVs. The GitHub repo remains linked from the body of the
README for contributors who want to submit results.下载附件