Spaces:
Sleeping
Sleeping
Commit History
-
Add reviewer overview figure, drop duplicate training SVGs, sync Hub showcase
luffy-orf- New reviewer_results_overview.svg (panels A–D) from build_reviewer_overview.py - Remove grpo_loss_kl, grpo_completion_lengths, grpo_reward_stability (covered by dashboard) - README/Blog link showcase without asking readers to run plot or upload scripts - refresh_plots.sh runs the new builder; showcase upload includes overview Made-with: Cursor2b6ff53 -
Shorten Hub dataset notes in README, Blog, and Hub README
luffy-orfMade-with: Cursor690c7b8 -
Tone down Hub dataset copy: neutral navigation, no trashing own plots
luffy-orfMade-with: Cursor0fff59e -
Hub dataset: root README with viewer disabled; document in README and Blog
luffy-orfPer Hugging Face docs, viewer:false stops the misleading default image viewer on mixed artifact repos; upload script now pushes that card with each showcase sync. Made-with: Cursor2fab1a1 -
Refresh plots, showcase upload script, and docs for Hub results
luffy-orfRegenerate eval and GRPO figures; prefer venv in refresh_plots.sh. Document that membrane-grpo-results updates via upload script, not git push. Made-with: Cursordbd3aaf -
submission: Membrane docs, training/eval artifacts, HF jobs scripts; docs plots as SVG only (HF Space git binary hook)
luffy-orfMade-with: Cursor9c49c6e