Remove results traces from model repo

Update checkpoint table to current 4 release models

Prune to 4 release checkpoints (v10, mo, gasa_generalist, camframe-74.3)

Add gasa_generalist (renamed gasa_E_box_50s)

Add full checkpoint set (all training runs)

Add all results.json archive

Add checkpoint swap_box_50s

Add checkpoint gasa_E_box_camframe_tpd_230s_50e

Add checkpoint gasa_E_box_camframe_230s_100ep_bs8

Add checkpoint gasa_E_box_world_230s_1008_converged

Add checkpoint gasa_E_box_50s

Add checkpoint ma_v10_config_245s_100ep

Update results: add UCO3D, LERF, NVOS, SPIn-NeRF benchmarks; add link to data repo

Update README.md

Update README.md

Upload v10/config.json with huggingface_hub

Upload v10/best.pt with huggingface_hub

Upload README.md with huggingface_hub

Delete fullscale_no_qp/config.json with huggingface_hub

Delete fullscale_no_qp/best.pt with huggingface_hub

Upload v10/best.pt with huggingface_hub

Upload v10/config.json with huggingface_hub

Fix param counts: DA3-NESTED-GIANT-LARGE is 1.69B, total 2.5B

Update model card: 230 scenes

Add fullscale_no_qp config

Add fullscale_no_qp checkpoint (single-object, 245s 100ep)

Add mo_v11 config

Add mo_v11 checkpoint (multi-object, text+spatial, 245s 8v 100ep)

Add model card

initial commit