VTWM E1 tactile drift: best vs last

Diffusion-forcing world model rollouts. Each sample is ~2.7 s (16 raw frames @ 6 fps). Side-by-side videos show ground truth on the left, model prediction on the right. First half of each clip = history (GT), second half = autoregressively sampled future.

Summary

Metrics are computed on saved future latent frames. Contact/no-contact split uses tactile latent-to-reference energy.

run	view MSE	TL MSE	TR MSE	tactile contact MSE	tactile no-contact MSE
E1_best_epoch36_val00200	0.049644	0.009406	0.004801	0.007103	n/a
E1_last_epoch99_val03365	0.049644	0.009406	0.004801	0.007103	n/a

E1_best_epoch36_val00200

E1_best_epoch36_val00200/sample_000

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_001

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_002

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_003

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_004

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_005

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_006

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_007

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_008

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_best_epoch36_val00200/sample_009

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365

E1_last_epoch99_val03365/sample_000

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_001

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_002

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_003

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_004

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_005

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_006

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_007

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_008

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout

E1_last_epoch99_val03365/sample_009

Panel (rows: tl, tr, view; cols: time, history|future split)

view_left — GT (L) vs rollout (R)

tactile_left — GT vs rollout

tactile_right — GT vs rollout