Rename checkpoint folders by benchmark suite

Remove internal wandb run-id files

Defer training-code release: remove training commands from cards

Sync model cards with GitHub release: paper-format results tables, spider plots, corrected citation

Link MODEL_CARD to GitHub repo + add reproduction snippet

Link MODEL_CARD to GitHub repo + add reproduction snippet

Link MODEL_CARD to GitHub repo + add reproduction snippet

Link MODEL_CARD to GitHub repo + add reproduction snippet

Link to companion GitHub repo + add HF→code quickstart

Update root README: swap Goal to v7+KL=0.1 @ 25k (97.8% STD), add Std column

Delete v7_kl015_goal_20k — replaced by v7_kl01_goal_25k (which has post-seeding 97.8% STD)

Add Anchor-Align v7+KL0.1 Goal 25k checkpoint (the 97.2/97.8 STD ckpt; post-seeding numbers)

Update root README with all 4 ckpts

Add Anchor-Align v7+KL0.15 LIBERO-10 45k checkpoint

Add Anchor-Align v7+KL0.15 Goal 20k checkpoint

Add Anchor-Align v7+KL0.15 Object 2.5k checkpoint

Add Anchor-Align v7+KL0.1 Spatial 10k checkpoint (3-seed evaluated)

Add repo-root README listing Spatial v7+KL0.1 ckpt

initial commit