WBench Examples: Reference Submissions for Interactive Video World Model Benchmark

Reference video submissions for WBench, a benchmark for interactive video world models. The dataset contains at least 447 videos across two model submissions, including 158 videos from HY-World 1.5 and 289 videos from Kling 3.0. It was uploaded by meituan-longcat and last updated on 2026-06-01.

Use Cases

• Benchmarking model performance based on the described 'camera' and 'text' interaction types
• Evaluating video generation consistency based on the per-video turn structure mentioned in the description
• Comparing different AI model architectures using the provided ready-to-evaluate submission packages

Strengths

• Contains at least 447 videos across two distinct model submissions
• Provides ready-to-evaluate submission packages structured for the WBench pipeline
• Includes metadata files (meta.json, turns.json) for each model submission as mentioned in the description

Limitations

• Column-level documentation is absent; field semantics must be inferred after download
• Row count is unknown, which may limit suitability assessment
• Description metadata is limited; actual data quality requires manual inspection after download

Provenance

Source

huggingface

Collection Method

Likely collected as reference submissions for the WBench benchmark evaluation pipeline.

Freshness

Last updated 2026-06-01 09:40:28