CapRL-Video-QA-20K: 20,000 Video Question-Answer Pairs

A subset of 20,000 video question-answer pairs from the LLaVA-Video-178K dataset, hosted by internlm. The dataset was last updated on 2026-05-22. It provides relative file paths to video clips, likely intended for training or evaluating multimodal AI models.

Use Cases

• Train video-based question-answering models based on the described QA pairs.
• Benchmark multimodal LLM performance on video comprehension tasks based on the provided video and question data.
• Fine-tune vision-language models for generating video captions or descriptions based on the associated QA format.

Strengths

• Derived from the larger LLaVA-Video-178K dataset, suggesting a foundation in a known multimodal corpus.
• Provides explicit relative video paths, facilitating data linkage.

Limitations

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

Provenance

Source

huggingface.co/datasets/internlm/CapRL-Video-QA-20K

Collection Method

Subset extracted from the lmms-lab/LLaVA-Video-178K dataset.

Freshness

Last updated 2026-05-22 13:29:20; freshness should be verified.