THVL/THVL-Bench
Dataset Card for Dataset Name THVL-Bench is a multimodal benchmark for Temporal Harmful Video Localization (THVL), designed to evaluate state-of-the-art Multimodal Large Language Models (MLLMs) on their ability to precisely localize harmful content in long-form videos, with fine-grained category labels and modality contribution annotations. The dataset contains 450 real-world long videos with 1,099 manually annotated harmful segments across 11 harm categories. This dataset card… See the full description on the dataset page: https://huggingface.co/datasets/THVL/THVL-Bench.
| Author / organization | THVL |
|---|---|
| License (as declared) | cc-by-4.0 |
| Downloads | 857 (as reported by Hugging Face) |
| Likes | 0 |
| Created | 7 May 2026 |
| Last updated at source | 5 August 2026 |
Popularity, last 2 days
857 now (+140 since 2026-10-16). Counts as reported by the source, snapshotted daily by Silytics.
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 17 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.