Trajectories









Bookmarks

Lowest advantage
episodes
(unexpected failures):

trajectory 6, frame 369
trajectory 5, frame 242
trajectory 2, frame 337
trajectory 6, frame 191
trajectory 8, frame 495
trajectory 7, frame 5
trajectory 1, frame 105
trajectory 7, frame 433
trajectory 5, frame 26
trajectory 4, frame 262
trajectory 8, frame 363
trajectory 7, frame 269
trajectory 3, frame 1
trajectory 4, frame 480
trajectory 8, frame 217
trajectory 3, frame 235

Highest advantage
episodes
(unexpected successes):

trajectory 2, frame 348
trajectory 5, frame 256
trajectory 8, frame 409
trajectory 4, frame 282
trajectory 7, frame 47
trajectory 6, frame 372
trajectory 8, frame 287
trajectory 5, frame 457
trajectory 6, frame 235
trajectory 7, frame 321
trajectory 6, frame 505
trajectory 8, frame 453
trajectory 5, frame 369
trajectory 8, frame 92
trajectory 4, frame 151
trajectory 3, frame 373

Layers














Timeline

frame: 1 policy: next action: A
no-op
A
B
fps
advantage
0.154
value function
9.60

Attribution

Observation Positive attribution Negative attribution

policy logits:

sums of policy logits:

Attribution legend

Click to expand feature

1
2
3
4
5
6
7
8

Hotkeys

go backwards
go forwards
toggle play/pause

Select a feature

Feature visualization

zoom in zoom out
fewer patches more patches