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.

Paper
Data
Model
EvalKit
Leaderboard

Data Engines

View All
dot_to_dot_task
GitHub
Knowledge in-domain testset
Prompt
The scene shows 8 numbered dots scattered across the image. Connect the dots in numerical order (1→2→3→...→8) by drawing red straight lines between them, one line at a time in sequence.
First Frame
First Frame
Last Frame
Last Frame
Video
shape_color_change
GitHub
Abstraction training set
Prompt
The scene shows an analogy A:B :: C:? with two rows of shapes and arrows. On the top row, a unknown heart becomes a unknown heart; treat this as the color rule from unknown to unknown. On the bottom row, the pentagon starts with the same source color. Apply the same color change so the answer shape on the right is a unknown pentagon, keeping its shape, size, and position the same while only the color changes.
First Frame
First Frame
Last Frame
Last Frame
Video
grid_obtaining_award
GitHub
Spatiality training set
Prompt
The scene shows a 10x10 grid with a green start point, a red end point, and 3 diamond reward items scattered across it. A circular agent starts at the green start point and can move to adjacent cells (up, down, left, right). The agent collects a reward by moving to its cell, and once collected, the reward disappears. Find the shortest path that collects all 3 diamond rewards before reaching the red end point.
First Frame
First Frame
Last 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
First Frame
Last Frame
Last Frame
Video
circle_all_squares_from_mixed_shapes
GitHub
Perception training set
Prompt
The scene shows 5 axis-aligned squares and rectangles. A square has equal width and height, while a rectangle does not. First identify all squares, then draw one red circle around each square. Do not circle any rectangles. Show the complete circling process step by step.
First Frame
First Frame
Last Frame
Last Frame
Video

Inference Results

View Full Bench
Identify Chinese Character - Samples
00
01
02
03
04
Task Domains 1/5
Identify Chinese Character
Knowledge out-of-domain testset
Shape Scale Then Outline
Abstraction in-domain testset
Grid Avoid Obstacles
Spatiality in-domain testset
Animal Matching
Transformation out-of-domain testset
Select Longest Side
Perception out-of-domain testset
Prompt
Loading...
Ground Truth
First
First Frame
Final
Final Frame
Model Outputs
1/
VBVR-Wan2.2
VBVR-Wan2.2
CogVideoX 1.5
Kling 2.6
LTX-2
Runway Gen-4
Sora 2
Veo 3
Wan 2.2 I2V
Hunyuan I2V
Seedance 2.0

Leaderboard

Modality
Split
Type
Category
2026-04-28