Categories Companies Research

Loading the SOTA2 catalog…

Explore

Categories Companies Research

Contribute

Join SOTA2 Contact

SOTA2

© 2026 SOTA2

Build in Bellevue, USA

WizWand is now SOTA2 Research

Papers Benchmarks Datasets Tasks

World Modeling research benchmarks – Page 2 · SOTA2 Research

World Modeling

Benchmarks

Dataset NameSOTA MethodSort

Dataset NameSOTA methodMetricTrendResultsLast Updated

River Raid (test)

Finite Automata Extraction

0.13FID

3

May 22, 2026

River Raid (train)

Finite Automata Extraction

0.14FID

3

May 22, 2026

Game Engine Learning

0.23FID

3

May 22, 2026

Pac-Man (train)

Game Engine Learning

0.19FID

3

May 22, 2026

MIND First 50 (test)

Matrix-Game 2.0

11.88Context Memory

3

Feb 26, 2026

MIND-Third 50 1.0 (test)

Matrix-Game 2.0

0.14Long Context Memory

3

Feb 26, 2026

RepairFlow Hybrid n=16

20HSR

2

Jun 29, 2026

17HSR

2

Jun 29, 2026

19.3HSR

2

Jun 29, 2026

14.2HSR

2

Jun 29, 2026

Robosuite Push Out-of-Distribution (test)

27.187PSNR (10 steps)

2

May 19, 2026

WorldModelBench Robot in office scenario (test)

Multi-Agent Framework

6.9Total Score

2

May 15, 2026

WorldModelBench Outdoor vehicle scenario (test)

Multi-Agent Framework

5.8Total Score

2

May 15, 2026

WorldModelBench Vehicle FSI scenario (test)

Multi-Agent Framework

6.8Total Score

2

May 15, 2026

WorldModelBench Robot in office

Multi-Agent Framework

2Instruction Following Score

2

May 15, 2026

WorldModelBench Outdoor vehicle

Multi-Agent Framework

1INSTR Score

2

May 15, 2026

WorldModelBench Vehicle FSI

Multi-Agent Framework

2.9Instruction Following Score

2

May 15, 2026

WorldModelBench Aggregated across three scenarios

Multi-Agent Framework

5.9Instruction Score

2

May 15, 2026

Page 2 of 2

Popular datasets

WorldModelBench MIND ScienceWorld WorldScore WorldArena

Follow for updates

Join SOTA2 Contact