WorldScore

WorldScore Icon WorldScore
A Unified Evaluation Benchmark for World Generation

Stanford University
^*Contributed Equally

Overview

The WorldScore benchmark unifies the evaluation of 3D, 4D, and video models on their ability to generate a world following instructions. Unlike existing benchmarks that focus on single-scene quality, we decompose world generation to a sequence of next-scene generation tasks based on explicit camera trajectories. Our WorldScore jointly measures controllability, quality, and dynamics.

Here we showcase how the WorldScore-Static metric measures two models given an initial scene of a bedroom with a specified camera path—"pan left" → "move left" → "pull out". While existing benchmarks rate Models A and B similarly based on single-scene video quality, our WorldScore benchmark differentiates their world generation capabilities by identifying that Model B fails to generate a new scene or follow the instructed camera movement.

Comparison of Benchmarks

Benchmark	# Examples	Multi-Scene	Unified	Long Seq.	Image Cond.	Multi-Style	Camera Ctrl.	3D Consist.
TC-Bench	150	❌	❌	❌	✅	❌	❌	❌
EvalCrafter	700	❌	❌	❌	❌	❌	❌	❌
FETV	619	❌	❌	❌	❌	❌	❌	❌
VBench	800	❌	❌	❌	❌	❌	❌	❌
T2V-CompBench	700	❌	❌	❌	❌	❌	❌	❌
Meng et al.	160	❌	❌	❌	❌	❌	❌	❌
Wang et al.	423	❌	❌	✅	❌	❌	❌	❌
ChronoMagic-Bench	1649	❌	❌	❌	❌	❌	❌	❌
WorldModelBench	350	❌	❌	❌	✅	❌	❌	❌
WorldScore (Ours)	3000	✅	✅	✅	✅	✅	✅	✅

Our WorldScore benchmark is designed to evaluate various world generation approaches including 3D, 4D, I2V and T2V models. It is designed to generate multiple scenes with varying sequence lengths. Our benchmark also features multiple visual styles, accurate camera control evaluation, and 3D consistency evaluation, all of which are important factors in world generation yet currently missing in existing benchmarks.

Controllability

Camera controllability evaluates how well the models adhere to the instructed layout. The layout instructions for each showcases are shown below the input image (e.g., "Camera pans right").

Camera pans right

Camera pans left

Camera pans right

Camera pushes in

Score: 99.96 Score: 0.00

Score: 99.90 Score: 0.00

Score: 99.97 Score: 0.00

Score: 99.95 Score: 0.00

A serene Urban Balcony featuring a lush potted plant...

School corridor, bulletin board, row of lockers; trophy display case; school courtyard, fountain, benches.

Score: 71.13 Score: 0.00

Score: 69.60 Score: 20.53

pancakes, orange juice

building, library, books, antique lamp

seating area, plants, city view, luxury lounge chairs

Score: 100.00 Score: 0.00

Score: 100.00 Score: 25.00

Quality

3D consistency focuses on how the geometry of a scene remains stable across frames, regardless of slight changes in visual textures.

Seaside Dining...

Urban Intersection...

A vintage truck...

Score: 92.88 Score: 0.00

Score: 94.60 Score: 0.00

Score: 94.41 Score: 0.00

A bright Modern Kitchen...

Clothing Store Interior...

Mountain Monastery...

Score: 100.00 Score: 25.27

Score: 63.94 Score: 0.00

Score: 92.77 Score: 36.94

Urban Rope Structure...

Urban Streetview...

Closed Escalator...

Score: 72.22 Score: 0.00

Score: 84.80 Score: 0.00

Score: 84.14 Score: 0.94

The cozy sofa and sleek coffee table are bathed in warm, natural light from the large windows.

Sunlight streams through large windows casting shadows over the statues and the checkered floor.

A breathtaking view of the Majestic Mountain Range as clouds gently caress the towering peaks.

In the Night Cityscape, the illuminated skyscrapers...

Score: 91.07 Score: 17.86

Score: 95.31 Score: 18.33

Score: 94.28 Score: 11.95

Score: 90.86 Score: 14.51

Dynamics

Motion accuracy measures whether the motion specified in the next-scene prompt (shown below the input image) occurs in the specified regions.

The octopus moves gracefully across the sandy ocean floor by extending and retracting its tentacles...

Water gently flows, creating subtle ripples on its surface.

Waterfall plunges down the cliff face, misting as it hits the pool. The stream flows swiftly over rocks...

The turtle slowly crawls forward into the water. The fish swims smoothly beneath the surface...

Score: 100.00 Score: 0.00

Score: 83.98 Score: 39.11

Score: 100.00 Score: 32.42

Score: 85.12 Score: 35.19

The panda gently shifts its weight from side to side as it climbs...

The large potted plant next to the desk sways gently side to side as if moved by a breeze...

The waterfall cascades down the cliff in a continuous, smooth flow...

The LEGO figure pivots and scans the horizon with its arm raised.

The car moves steadily across the snow-covered terrain, leaving deep tracks behind it as it navigates icy landscape.

Score: 90.44 Score: 13.92

Score: 92.21 Score: 13.42

Score: 91.44 Score: 20.24

Score: 91.06 Score: 20.54

Score: 46.89 Score: 18.91

The sea turtle glides smoothly through the water, propelling itself forward with slow, graceful movements...

The rocket blasts upwards into the sky, propelled by powerful thrust from its engines. The clouds of smoke billow outward rapidly...

The octopus glides smoothly through the shallow water, its tentacles undulating gently to propel it forward.

Score: 100.00 Score: 37.84

Score: 100.00 Score: 28.57

Score: 100.00 Score: 29.80

Abstract

We introduce the WorldScore benchmark, the first unified benchmark for world generation. We decompose world generation into a sequence of next-scene generation tasks with explicit camera trajectory-based layout specifications, enabling unified evaluation of diverse approaches from 3D and 4D scene generation to video generation models. The WorldScore benchmark encompasses a curated dataset of 3,000 test examples that span diverse worlds: static and dynamic, indoor and outdoor, photorealistic and stylized. The WorldScore metric evaluates generated worlds through three key aspects: controllability, quality, and dynamics. Through extensive evaluation of 19 representative models, including both open-source and closed-source ones, we reveal key insights and challenges for each category of models.

Evaluation Results

WorldScore-Static aggregates controllability and quality. WorldScore-Dynamic aggregates controllability, quality, and dynamics. See our

for the latest ranking and numerical results.

Models	WorldScore-Static	WorldScore-Dynamic	Camera Ctrl	Object Ctrl	Content Align	3D Consist	Photo Consist	Style Consist	Subjective Qual	Motion Acc	Motion Mag	Motion Smooth
Gen-3	60.71	57.58	29.47	62.92	50.49	68.31	87.09	62.82	63.85	54.53	27.48	68.87
Hailuo	57.55	56.36	22.39	69.56	73.53	67.18	62.82	54.91	52.44	63.46	27.20	70.07
DynamiCrafter	52.09	47.19	25.15	47.36	25.00	72.90	60.95	78.85	54.40	41.11	39.25	26.92
VideoCrafter1-T2V	47.10	43.54	21.61	50.44	60.78	64.86	51.36	38.05	42.63	11.76	75.00	18.87
VideoCrafter1-I2V	50.47	47.64	25.46	24.25	35.27	74.42	73.89	65.17	54.85	55.63	25.00	42.49
VideoCrafter2	52.57	47.49	28.92	39.07	72.46	65.14	61.85	43.79	56.74	47.12	30.40	29.39
T2V-Turbo	45.65	40.20	27.80	30.68	69.14	38.72	34.84	49.65	68.74	34.87	40.09	7.48
EasyAnimate	52.85	51.65	26.72	54.50	50.76	67.29	47.35	73.05	50.31	75.00	31.16	40.32
CogVideoX-T2V	54.18	48.79	40.22	51.05	68.12	68.81	64.20	42.19	44.67	25.00	47.31	36.28
CogVideoX-I2V	62.15	59.12	38.27	40.07	36.73	86.21	88.12	83.22	62.44	69.56	26.42	60.15
Allegro	55.31	51.97	24.84	57.47	51.48	70.50	69.89	65.60	47.41	54.39	40.28	37.81
Vchitect-2.0	42.28	38.47	26.55	49.54	65.75	41.53	42.30	25.69	44.58	33.59	33.81	21.31
SceneScape	50.73	35.51	84.99	47.44	28.64	76.54	62.88	21.85	32.75	0.00	0.00	0.00
Text2Room	62.10	43.47	94.01	38.93	50.79	88.71	88.36	37.23	36.69	0.00	0.00	0.00
LucidDreamer	70.40	49.28	88.93	41.18	75.00	90.37	90.20	48.10	58.99	0.00	0.00	0.00
WonderJourney	63.75	44.63	84.60	37.10	35.54	80.60	79.03	62.82	66.56	0.00	0.00	0.00
InvisibleStitch	61.12	42.78	93.20	36.51	29.53	88.51	89.19	32.37	58.50	0.00	0.00	0.00
WonderWorld	72.69	50.88	92.98	51.76	71.25	86.87	85.56	70.57	49.81	0.00	0.00	0.00
4D-fy	27.98	32.10	69.92	55.09	0.85	35.47	1.59	32.04	0.89	22.22	22.88	80.06