LiveSports-3K is a benchmark for evaluating streaming video understanding capabilities of large language and multimodal models. It consists of two evaluation tracks, including a Closed Captions track that measures real-time commentary generation aligned with ground-truth ASR transcripts. The dataset was created by stdKonjac and updated on May 20, 2025.
Use Cases
- Benchmarking model performance on real-time commentary generation based on the Closed Captions track
- Evaluating multimodal model alignment with ASR transcripts based on the described ground-truth
- Testing streaming video understanding capabilities based on the benchmark's stated purpose
Strengths
- Designed specifically for evaluating streaming video understanding, a defined task
- Includes a Closed Captions track with ground-truth ASR transcripts, a concrete feature
- Updated on May 20, 2025, indicating recent maintenance
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
- Freshness
- Last updated 2025-05-20 16:46:16