Dataset · Hugging Face

sebastian-stapf/WorldTasks

Name: WorldTasks
Creator: sebastian-stapf
License: other
Keywords: task_categories:text-to-video, task_categories:image-to-video, language:en, license:other, size_categories:10K<n<100K, format:imagefolder, modality:image, library:datasets, library:mlcroissant, region:us, world-models, self-distillation, task-conditioned-video-generation, video-generation

WorldTasks WorldTasks is the task-conditioned image/video world-model dataset for World Model Self-Distillation: Training World Models to Solve General Tasks. WorldTasks pairs visual scenes with compact task instructions and, for training records, detailed solution descriptions. The dataset is designed to turn image/video generators into task-conditioned world models: a model sees an initial scene and a short instruction, then learns to synthesize a plausible task completion.… See the full description on the dataset page: https://huggingface.co/datasets/sebastian-stapf/WorldTasks.

View on Hugging FaceReport an issue

Author / organization	sebastian-stapf
License (as declared)	other
Downloads	242 (as reported by Hugging Face)
Likes	2
Created	3 June 2026
Last updated at source	11 June 2026

world-modelsself-distillationtask-conditioned-video-generationvideo-generation

Embed this entry

<iframe src="https://silytics.si/embed/item/2ae75fd1-b6ac-4eb9-805a-d28e0940edca" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Free, no tracking. More about feeds and embeds.

Popularity, last 6 days

242 now (+45 since 2026-10-21). Counts as reported by the source, snapshotted daily by Silytics.

More from sebastian-stapf

sebastian-stapf/action-progress-droidDataset
sebastian-stapf/action-progress-videoDataset

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 26 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.