WorldTasks: Task-Conditioned Image and Video Data for World Model Training

WorldTasks is a dataset for training world models via self-distillation, created by sebastian-stapf and hosted on Hugging Face. It pairs visual scenes with compact task instructions and detailed solution descriptions. The dataset was last updated on June 11, 2026.

Use Cases

• Training task-conditioned world models based on paired visual scenes and instructions.
• Self-distillation of world models based on solution descriptions for training records.
• Synthesizing plausible video completions based on an initial scene and a short instruction.

Strengths

• Designed for a specific research purpose: turning image/video generators into task-conditioned world models.
• Pairs multiple data modalities: visual scenes, task instructions, and solution descriptions.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and dataset size are unknown, which may limit suitability assessment.
• License information is unknown, which may restrict usage.

Provenance

Source

sebastian-stapf on Hugging Face

Freshness

Last updated 2026-06-11 19:03:33; freshness should be verified.