Dataset Viewer
Auto-converted to Parquet Duplicate
scene_id
int64
slot
int64
donor
video
receiver
video
donor_label
string
receiver_label
string
membership
list
donor_clip_id
string
donor_crop_index
int64
donor_width
int64
donor_height
int64
donor_sample_frames
list
receiver_clip_id
string
receiver_crop_index
int64
receiver_width
int64
receiver_height
int64
receiver_sample_frames
list
0
1
bird
airplane
[ "train", "bird", "clock" ]
commons-b05
0
852
480
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
airplane-ups
1
1,574
702
[ 12, 36, 60, 84, 108, 132, 156, 180 ]
1
6
train
cat
[ "laptop", "train", "horse" ]
train-viaduct
1
1,920
1,080
[ 2, 7, 12, 17, 23, 28, 33, 38 ]
commons-ca10
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
2
1
cat
airplane
[ "horse", "airplane", "laptop" ]
commons-ca10
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
airplane-lake
3
522
306
[ 7, 22, 37, 52, 67, 82, 97, 112 ]
3
5
airplane
cat
[ "cat", "dog", "cup" ]
airplane-lake
3
522
306
[ 7, 22, 37, 52, 67, 82, 97, 112 ]
commons-c02
4
854
480
[ 67, 202, 337, 472, 607, 742, 877, 1012 ]
4
3
dog
bicycle
[ "dog", "bird", "car" ]
commons-d02
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
bike-tandem
5
1,170
572
[ 5, 16, 27, 38, 50, 61, 72, 83 ]
5
5
chair
horse
[ "clock", "boat", "chair" ]
uco3d-chair-63738-33143-53795
5
1,080
1,600
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h12
6
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
6
2
bird
airplane
[ "boat", "airplane", "laptop" ]
commons-b05
0
852
480
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
airplane-an12
7
1,478
714
[ 25, 76, 127, 178, 228, 279, 330, 381 ]
7
1
chair
horse
[ "car", "horse", "bicycle" ]
uco3d-chair-13693-50456-50065
0
1,080
1,650
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h17
8
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
8
4
cat
bicycle
[ "airplane", "cat", "cup" ]
commons-c11
8
1,280
720
[ 30, 91, 153, 214, 275, 336, 398, 459 ]
bike-karelian
9
1,920
1,080
[ 2, 8, 13, 18, 24, 29, 34, 40 ]
9
1
laptop
horse
[ "airplane", "clock", "laptop" ]
uco3d-laptop-179-30684-92228
0
1,400
914
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h15
10
1,280
720
[ 5, 17, 29, 41, 53, 65, 77, 89 ]
10
6
dog
bus
[ "cat", "bus", "bicycle" ]
commons-da09
1
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
bus-airport
1
1,920
1,080
[ 3, 11, 19, 27, 34, 42, 50, 58 ]
11
3
airplane
dog
[ "cat", "boat", "dog" ]
airplane-a380
0
1,920
1,080
[ 27, 81, 135, 189, 243, 297, 351, 405 ]
commons-d01
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
12
5
horse
cup
[ "car", "bus", "horse" ]
commons-h01
1
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-cup-latte-art
1
620
380
[ 16, 49, 82, 115, 148, 181, 214, 247 ]
13
6
chair
horse
[ "bus", "train", "chair" ]
uco3d-chair-51765-64499-61598
6
1,080
1,710
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h09
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
14
3
horse
laptop
[ "laptop", "bus", "bird" ]
commons-h09
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
uco3d-laptop-133-49764-51686
6
946
562
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
15
5
bicycle
dog
[ "airplane", "dog", "boat" ]
bike-person04
3
1,024
450
[ 2, 6, 10, 15, 19, 24, 28, 32 ]
commons-d11
7
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
16
5
bird
car
[ "horse", "bird", "airplane" ]
commons-b06
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
car-ferrari
2
1,272
474
[ 0, 2, 4, 5, 7, 8, 10, 12 ]
17
2
bus
cat
[ "car", "bus", "clock" ]
bus-capacity
7
1,364
668
[ 2, 7, 12, 17, 22, 27, 32, 37 ]
commons-c05
7
1,920
1,080
[ 15, 47, 79, 110, 142, 173, 205, 237 ]
18
4
bird
train
[ "cat", "horse", "train" ]
commons-b05
0
852
480
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
train-cnr
8
1,920
1,080
[ 8, 25, 42, 59, 77, 94, 111, 128 ]
19
2
train
cat
[ "bird", "pizza", "cat" ]
train-cnr
8
1,920
1,080
[ 8, 25, 42, 59, 77, 94, 111, 128 ]
commons-ca09
9
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
20
6
bird
airplane
[ "clock", "cup", "bird" ]
commons-b04
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
plane-lviv
10
1,272
496
[ 12, 36, 60, 84, 108, 132, 156, 180 ]
21
6
boat
dog
[ "pizza", "boat", "bird" ]
boat-manila
3
1,280
720
[ 13, 40, 67, 94, 121, 148, 175, 202 ]
commons-d02
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
22
3
horse
pizza
[ "cup", "pizza", "boat" ]
commons-h16
0
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-pizza-trento
1
1,080
1,080
[ 24, 72, 120, 168, 216, 264, 312, 360 ]
23
1
laptop
horse
[ "clock", "horse", "chair" ]
uco3d-laptop-62078-62954-13367
5
1,048
634
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h02
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
24
2
cat
bus
[ "airplane", "cat", "train" ]
commons-ca10
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
bus-ltz1130
5
1,280
720
[ 1, 3, 6, 8, 11, 13, 16, 18 ]
25
3
train
dog
[ "train", "cup", "laptop" ]
train-amtrak
3
1,280
720
[ 4, 12, 20, 29, 37, 46, 54, 62 ]
commons-da04
8
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
26
3
bird
airplane
[ "pizza", "dog", "airplane" ]
commons-b04
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
airplane-il76
5
1,920
1,080
[ 15, 45, 75, 105, 135, 165, 195, 225 ]
27
3
boat
dog
[ "bus", "dog", "horse" ]
boat-local
0
1,920
1,080
[ 7, 22, 37, 52, 67, 82, 97, 112 ]
commons-da09
1
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
28
3
cat
bicycle
[ "laptop", "car", "cat" ]
commons-c03
6
1,920
1,080
[ 5, 15, 25, 35, 46, 56, 66, 76 ]
bike-tandem
5
1,170
572
[ 5, 16, 27, 38, 50, 61, 72, 83 ]
29
5
car
dog
[ "bird", "car", "boat" ]
car-rally-loeb
4
1,920
1,080
[ 1, 4, 7, 10, 14, 17, 20, 23 ]
commons-d01
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
30
6
horse
clock
[ "train", "boat", "clock" ]
commons-h17
8
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-comtoise
7
600
1,080
[ 9, 27, 45, 63, 81, 99, 117, 135 ]
31
2
bus
dog
[ "cup", "horse", "dog" ]
bus-airport
1
1,920
1,080
[ 3, 11, 19, 27, 34, 42, 50, 58 ]
commons-d08
5
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
32
4
cat
boat
[ "cat", "laptop", "cup" ]
commons-ca02
10
1,280
720
[ 5, 17, 29, 41, 53, 65, 77, 89 ]
boat-columbia
6
1,306
1,080
[ 9, 27, 46, 65, 83, 102, 121, 139 ]
33
2
car
dog
[ "pizza", "car", "chair" ]
car-rally-guillaume
3
1,920
1,080
[ 3, 9, 15, 22, 28, 35, 41, 47 ]
commons-d11
7
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
34
2
bird
train
[ "train", "laptop", "chair" ]
commons-b06
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
train-rome
2
1,280
720
[ 6, 19, 32, 45, 57, 70, 83, 96 ]
35
4
bicycle
cat
[ "laptop", "cat", "car" ]
bike-west-virginia
11
1,584
600
[ 1, 3, 5, 7, 9, 11, 13, 15 ]
commons-c01
3
1,920
1,080
[ 12, 37, 62, 87, 113, 138, 163, 188 ]
36
6
horse
clock
[ "laptop", "horse", "bicycle" ]
commons-h09
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-reverse
1
1,080
1,080
[ 24, 72, 120, 168, 216, 264, 312, 360 ]
37
4
airplane
bird
[ "airplane", "chair", "cat" ]
airplane-ezy
4
1,920
1,080
[ 11, 35, 59, 83, 107, 131, 155, 179 ]
commons-b04
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
38
1
dog
train
[ "clock", "train", "car" ]
commons-da01
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
train-rovigo
6
1,280
720
[ 11, 35, 59, 83, 107, 131, 155, 179 ]
39
4
car
bird
[ "chair", "bird", "clock" ]
car-rally-loeb
4
1,920
1,080
[ 1, 4, 7, 10, 14, 17, 20, 23 ]
commons-b06
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
40
6
horse
pizza
[ "horse", "car", "cat" ]
commons-h14
7
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-pizza-bubbling
8
1,420
900
[ 6, 19, 31, 44, 57, 70, 82, 95 ]
41
3
bus
cat
[ "bus", "train", "bird" ]
bus-airport
1
1,920
1,080
[ 3, 11, 19, 27, 34, 42, 50, 58 ]
commons-ca09
9
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
42
2
horse
clock
[ "clock", "bicycle", "chair" ]
commons-h18
9
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-comtoise
7
600
1,080
[ 9, 27, 45, 63, 81, 99, 117, 135 ]
43
1
laptop
horse
[ "clock", "bird", "horse" ]
uco3d-laptop-59538-30649-34468
7
994
1,228
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h16
0
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
44
3
cat
bus
[ "train", "boat", "cat" ]
commons-ca03
0
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
bus-ltz1130
5
1,280
720
[ 1, 3, 6, 8, 11, 13, 16, 18 ]
45
5
boat
cat
[ "boat", "bus", "chair" ]
boat-local
0
1,920
1,080
[ 7, 22, 37, 52, 67, 82, 97, 112 ]
commons-ca10
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
46
4
dog
boat
[ "airplane", "laptop", "boat" ]
commons-da09
1
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
boat-burullus
2
1,920
1,080
[ 5, 15, 25, 36, 46, 57, 67, 77 ]
47
1
bicycle
dog
[ "dog", "boat", "bus" ]
bike-west-virginia
11
1,584
600
[ 1, 3, 5, 7, 9, 11, 13, 15 ]
commons-d01
3
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
48
1
bird
train
[ "horse", "chair", "bird" ]
commons-b05
0
852
480
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
train-tobu
5
1,920
1,080
[ 3, 10, 17, 24, 30, 37, 44, 51 ]
49
2
clock
horse
[ "chair", "car", "clock" ]
commons-reverse
1
1,080
1,080
[ 24, 72, 120, 168, 216, 264, 312, 360 ]
commons-h12
6
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
50
5
bird
car
[ "car", "clock", "bus" ]
commons-b04
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
car-landcruiser
1
1,166
662
[ 3, 11, 18, 26, 33, 41, 48, 56 ]
51
4
bus
cat
[ "cup", "bird", "cat" ]
bus-airport
1
1,920
1,080
[ 3, 11, 19, 27, 34, 42, 50, 58 ]
commons-c11
8
1,280
720
[ 30, 91, 153, 214, 275, 336, 398, 459 ]
52
1
dog
boat
[ "train", "dog", "car" ]
commons-d11
7
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
boat-birdseye
8
1,920
1,080
[ 7, 22, 37, 52, 67, 82, 97, 112 ]
53
2
bicycle
cat
[ "laptop", "bicycle", "cup" ]
bike-tandem
5
1,170
572
[ 5, 16, 27, 38, 50, 61, 72, 83 ]
commons-ca02
10
1,280
720
[ 5, 17, 29, 41, 53, 65, 77, 89 ]
54
5
horse
cup
[ "clock", "dog", "cup" ]
commons-h15
10
1,280
720
[ 5, 17, 29, 41, 53, 65, 77, 89 ]
commons-cup-coffee-sugar
3
1,452
1,080
[ 25, 76, 126, 177, 228, 279, 329, 380 ]
55
2
cup
horse
[ "dog", "car", "horse" ]
commons-cup-coffee-sugar
3
1,452
1,080
[ 25, 76, 126, 177, 228, 279, 329, 380 ]
commons-h01
1
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
56
2
bird
car
[ "pizza", "bird", "chair" ]
commons-b04
2
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
car-ferrari
2
1,272
474
[ 0, 2, 4, 5, 7, 8, 10, 12 ]
57
4
train
cat
[ "horse", "chair", "train" ]
train-rome
2
1,280
720
[ 6, 19, 32, 45, 57, 70, 83, 96 ]
commons-c01
3
1,920
1,080
[ 12, 37, 62, 87, 113, 138, 163, 188 ]
58
1
bird
car
[ "train", "bicycle", "car" ]
commons-b06
4
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
car-rally-loeb
4
1,920
1,080
[ 1, 4, 7, 10, 14, 17, 20, 23 ]
59
1
car
dog
[ "dog", "airplane", "cat" ]
car-rally-loeb
4
1,920
1,080
[ 1, 4, 7, 10, 14, 17, 20, 23 ]
commons-d09
6
1,280
720
[ 8, 24, 40, 57, 73, 90, 106, 122 ]
60
5
dog
boat
[ "bus", "dog", "chair" ]
commons-d09
6
1,280
720
[ 8, 24, 40, 57, 73, 90, 106, 122 ]
boat-stockholm
7
1,286
962
[ 18, 54, 90, 126, 162, 198, 234, 270 ]
61
1
chair
horse
[ "pizza", "laptop", "chair" ]
uco3d-chair-63738-33143-53795
5
1,080
1,600
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
commons-h14
7
1,920
1,080
[ 6, 18, 30, 42, 54, 66, 78, 90 ]
62
6
cat
train
[ "train", "horse", "boat" ]
commons-c05
7
1,920
1,080
[ 15, 47, 79, 110, 142, 173, 205, 237 ]
train-cnr
8
1,920
1,080
[ 8, 25, 42, 59, 77, 94, 111, 128 ]
63
6
clock
horse
[ "bird", "horse", "dog" ]
commons-clock-huber
3
1,440
1,440
[ 10, 31, 52, 73, 94, 115, 136, 157 ]
commons-h18
9
1,280
720
[ 6, 18, 30, 42, 54, 66, 78, 90 ]

Jspace Matched Video

This dataset pairs short, real video clips for the video version of the Jspace image activation-patching experiment. The test split has 64 scene rows and 73 unique videos. Each row contains a playable donor and receiver, their labels, the original numbered slot, and the three labels used in the membership question. Video reuse follows the original image crop schedule.

from datasets import load_dataset

scenes = load_dataset("saxenan3/jspace-matched-video", split="test")

Use in the experiment

Each video has eight fixed sample indices in the metadata. Render both sides in the specified slot using the existing Jspace canvas. Use the same middle sampled frame for the native still and the repeated-frame video control. render_inputs.py prepares these three inputs for one scene in the Jspace environment:

hf download saxenan3/jspace-matched-video --repo-type dataset --local-dir jspace-video
uv run python jspace-video/render_inputs.py --data-dir jspace-video --scene 0 --out scene-000.npz

Use the image experiment's receiver prompts and four-layer residual patch windows. Capture the donor under the neutral scene prompt. Run the model's clean naming, self-patch, and full donor patch controls before interpreting a sweep. This release supplies stimuli and the pair schedule. Model gate outcomes are not included.

Provenance and checks

All clips originate in real recordings from Wikimedia Commons or Meta AI's uCO3D. The released MP4s are silent H.264 cuts at 24 fps. Selected views are cropped where needed. Original capture resolution and the source and output dimensions are recorded in ATTRIBUTION.tsv, with the creator, original item or archive member, source interval, applicable license, and SHA-256 hash for each clip. Individual source licenses apply to the corresponding video.

The cuts were screened in sampled frames. Each final video was decoded at its eight published sample indices, and the two video columns were checked against the frozen 64-row schedule. Camera movement supplies the changing view in some household clips. Clips vary in source resolution, duration, context, and framing. This small, curated set is intended for the matched intervention experiment.

Downloads last month
28