WorldModelBenchmark/playworld-bench-anonymous

PlayWorld: Benchmarking World Models with Agent Players over Long-Horizon Objectives Anonymous benchmark release for double-blind review. Author, affiliation, and identity-bearing links are intentionally omitted. Project | Code Video world models generate future states from an initial observation and user actions. Comparing interactive models fairly is difficult because the actions required to reach the same long-horizon objective can vary substantially across models. PlayWorld… See the full description on the dataset page: https://huggingface.co/datasets/WorldModelBenchmark/playworld-bench-anonymous.