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
identify_one_and_nine
GitHub
Knowledge training set
The image shows a subset of digits chosen from 1 to 9 placed in different positions. Find digit 1 and digit 9. Only circle digits '1' and '9'. Do not circle other digits. Draw a red circle around each target digit.
First Frame
Last Frame
control_panel
GitHub
Abstraction out-of-domain testset
The image shows a control panel with six identical control units. Each unit has a colored indicator light at the top and a control lever at the bottom that can be moved to three positions (left, middle, or right). Observe the current control panel to infer the relationship between lever positions and light colors. Based on this inferred relationship, adjust the levers that need to be changed to make all indicator lights show yellow color.
First Frame
Last Frame
select_leftmost_shape
GitHub
Spatiality out-of-domain testset
Multiple shapes are shown. Circle the leftmost one. Do not change anything else.
First Frame
Last Frame
2d_object_rotation
GitHub
Transformation out-of-domain testset
The scene contains 4 2D object(s). Show them rotating clockwise by 53 degrees around their respective centroids.
First Frame
Last Frame
select_longest_polygon_side
GitHub
Perception out-of-domain testset
The image shows an irregular polygon with 6 sides. First compare the lengths of all polygon edges, then mark the single longest side by drawing a small circle at its midpoint. Do not change anything else. Show the complete solution step by step.
First Frame
Last Frame

Inference Results

View Full Bench
Traffic Light - Samples
00
01
02
03
04
Task Domains 1/5
Traffic Light
Knowledge in-domain testset
Shape Outline Then Move
Abstraction in-domain testset
LEGO Construction
Spatiality in-domain testset
Shape Sorter
Transformation out-of-domain testset
Mark Second Largest Shape
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