Link MuJoCo sim twin dataset

superseded by grasp-discrete-{25,15}

card: 540 eps, distractor-jitter tranche, highlight-style tranche figures

regenerate grasp-twist-cw.png (highlight style, 540 eps)

regenerate grasp-twist-ccw.png (highlight style, 540 eps)

regenerate grasp-distractor.png (highlight style, 540 eps)

regenerate grasp-discrete-25.png (highlight style, 540 eps)

regenerate grasp-discrete-15.png (highlight style, 540 eps)

regenerate grasp-dense.png (highlight style, 540 eps)

regenerate grasp-blanket.png (highlight style, 540 eps)

regenerate grasp-all.png (highlight style, 540 eps)

add newly recorded varied episodes (pre 2.56M b96 train)

Point model link at renamed pick-cube-smolvla-so101

Link the finetuned SmolVLA model

Regenerate all grasp figures (highlight vs 440; twist-CC->CCW + new twist-CW); HTML image table

Card: 440 eps, twist-CC->CCW, add twist-CW; drop internal notes

add twist-CW grasp figure

grasp-all -> 440 eps

Add twist-CW jitter tranche (eps 340-439); 340->440 episodes

drop superseded scatter figures

freeform dataset card

add grasp-distribution figures

Upload README.md with huggingface_hub

Upload media/scatter-all.png with huggingface_hub

Upload README.md with huggingface_hub

Upload folder using huggingface_hub

Upload README.md with huggingface_hub

Upload media/scatter-continuous.png with huggingface_hub

Upload media/scatter-discrete.png with huggingface_hub

Upload media/scatter-all.png with huggingface_hub

Upload README.md with huggingface_hub

Upload folder using huggingface_hub

initial commit