A Very Big Video Reasoning Suite

We bet on a future that video reasoning is the next fundamental intelligence paradigm, after language reasoning, where spatiotemporal embodied world experiences could be more naturally captured.

Data Engines

VIEW ALL DATA ENGINE

clock

GitHub

Knowledge in-domain testset

Prompt

The clock shows 6:53. Show what the clock will look like after 2 hours.

First Frame

Last Frame

Video

symmetry_shape

GitHub

Abstraction training set

Prompt

The scene shows a grid with a continuous shape on the left half. Expand the shape symmetrically to the right half by mirroring it across the vertical axis, creating a complete symmetric shape.

First Frame

Last Frame

Video

grid_go_through_block

GitHub

Spatiality in-domain testset

Prompt

The scene shows a 10x10 grid with a green start square (containing an orange circular agent), a red end square, and multiple blue, purple and pink rectangular blocks. Starting from the green start square, the agent can move to adjacent cells (up, down, left, right). The goal is to move the agent to the red end square along the shortest path that passes through all blue, purple and pink blocks (the agent must visit every blue, purple and pink block before reaching the red end square).

First Frame

Last Frame

Video

combined_objects_spinning

GitHub

Transformation training set

Prompt

The scene shows 2 objects on the left side and dashed target outlines on the right side. The dashed target outlines remain completely stationary. For each object, first rotate it in place to match the orientation of its corresponding dashed target outline, then move it horizontally to the right so that it aligns exactly with and fits within its corresponding dashed target outline.

First Frame

Last Frame

Video

locate_intersection_of_segments

GitHub

Perception out-of-domain testset

Prompt

The scene shows two black line segments that intersect at exactly one point. First locate the single intersection point where the two segments cross, then draw one red circle around that intersection point. Do not circle any endpoints and do not add any extra lines. Show the complete marking process step by step.

First Frame

Last Frame

Video

Inference Results

View All Results

Dot to Dot - Samples

Task Domains 1/5

Dot to Dot

Knowledge in-domain testset

Shape Outline Then Move

Abstraction in-domain testset

Outline Innermost Square

Spatiality out-of-domain testset

Rolling Ball

Transformation in-domain testset

Color Triple Intersection

Perception out-of-domain testset

Prompt

Ground Truth

First

Final

Model Outputs

1/9

VBVR-Wan2.2

CogVideoX 1.5

Kling 2.6

LTX-2

Runway Gen-4

Sora 2

Veo 3

Wan 2.2 I2V

Hunyuan I2V

Leaderboard

Split