Drop Upload-your-own tab

Add paper-baselines comparison table

F1 fix

Filter benchmark labels to 7-class subset

Re-add Benchmark-your-model tab

Re-apply schema-walker monkey-patch

Restore pre-BYOM Space

Demote BYOM tab to CLI docs

Bare launch()

Patch gradio_client schema walker

launch show_api=False bypass

Revert gradio + strip schema-breaking kwargs

Add Benchmark-your-model tab

Initial Space: Gradio demo, weights pulled from Pybunny/nilmbench-faustine