Meta tags:
description=An unprecedentedly large-scale suite for video reasoning, comprising VBVR-Dataset, VBVR-Bench, and VBVR-DataFactory, enabling systematic scaling studies and emergent generalization in video reasoning models.;
author=VBVR Team;

Headings (most frequently used words):

data, very, big, video, reasoning, suite, engines, inference, results, leaderboard, resources, models,

Text of the page (most frequently used words):
the (28), video (14), vbvr (12), domain (12), testset (10), frame (10), and (8), #reasoning (7), data (7), prompt (7), first (6), bench (5), all (5), out (5), generate (5), another (5), last (5), with (5), github (5), very (4), big (4), suite (4), wan2 (4), image (4), red (4), shows (4), maze (4), position (4), preview (3), leaderboard (3), model (3), identify (3), that (3), scene (3), dashed (3), target (3), right (3), green (3), through (3), average (3), temperature (3), clock (3), 2026 (2), dataset (2), datafactory (2), thinkmorph (2), bagel (2), outputs (2), ground (2), truth (2), loading (2), i2v (2), perception (2), transformation (2), spatiality (2), abstraction (2), knowledge (2), chinese (2), character (2), view (2), inference (2), results (2), exactly (2), asymmetrical (2), circle (2), left (2), side (2), outlines (2), move (2), each (2), corresponding (2), dark (2), walls (2), white (2), pathways (2), starting (2), flag (2), end (2), from (2), start (2), find (2), axis (2), point (2), after (2), engine (2), engines (2), models, resources, category, proprietary, open, source, type, split, modality, nano, banana, gpt, sensenova, seedance, hunyuan, wan, veo, sora, runway, gen, ltx, kling, cogvideox, final, spot, unique, color, rolling, ball, key, door, matching, construction, blueprint, task, domains, samples, full, among, displayed, shapes, one, shape, not, change, anything, else, mark_asymmetrical_shape, objects, remain, completely, stationary, object, horizontally, aligns, fits, within, its, outline, separate_objects_no_spin, grid, circular, marker, indicates, marks, navigate, moving, along, you, can, adjacent, cells, down, but, cannot, pass, goal, demonstrate, complete, path, showing, step, journey, line, chart, titled, monthly, beijing, 2024, month, lowest, draw, rectangular, border, around, highlight, chart_extreme_with_data, show, what, will, look, like, hours, evalkit, paper, bet, future, next, fundamental, intelligence, paradigm, language, where, spatiotemporal, embodied, world, experiences, could, more, naturally, captured, demystifying, projects, skip, main, content,

Text of the page (random words):
a very big video reasoning suite skip to main content data data engines vbvr datafactory vbvr dataset bench inference results vbvr bench leaderboard projects vbvr wan2 2 demystifying video reasoning 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 video data engine image data engine preview view all 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 generate another chart_extreme_with_data github abstraction in domain testset prompt the scene shows a line chart titled monthly average temperature in beijing 2024 with month on the x axis and average temperature on the y axis find the point with the lowest average temperature and draw a red rectangular border around the corresponding point to highlight it first frame last frame video generate another maze github spatiality out of domain testset prompt the scene shows a 15 15 grid maze with dark walls and white pathways a green circular marker indicates the starting position and a red flag marks the end position starting from the green start position navigate through the maze by moving along the white pathways you can move to adjacent cells up down left right but cannot pass through the dark walls the goal is to find and demonstrate the complete path from the green start to the red flag end position showing each step of the journey through the maze first frame last frame video generate another separate_objects_no_spin github transformation out of domain testset 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 move each object horizontally to the right so that it aligns exactly with and fits within its corresponding dashed target outline first frame last frame video generate another mark_asymmetrical_shape github perception out of domain testset prompt among the displayed shapes exactly one is asymmetrical identify and circle that asymmetrical shape with a red circle do not change anything else first frame last frame video generate another inference results video bench image preview view full bench identify chinese character samples 00 01 02 03 04 task domains 1 5 identify chinese character knowledge out of domain testset construction blueprint abstraction in domain testset key door matching spatiality in domain testset rolling ball transformation in domain testset spot unique color perception in domain testset prompt loading ground truth first final 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 prompt loading ground truth model outputs 1 vbvr bagel bagel sensenova u1 vbvr thinkmorph thinkmorph gpt image 2 nano banana leaderboard modality all video image preview split all in domain out of domain type all open source proprietary category all 2026 04 28 a very big video reasoning suite resources vbvr bench vbvr datafactory data models vbvr dataset vbvr wan2 2 2026 a very big video reasoning suite