VideoChat2-IT-clean: A Cleaned Video Instruction Tuning Dataset

VideoChat2-IT-clean is a cleaned version of the VideoChat2-IT video instruction tuning dataset, released alongside the ICLR 2026 paper 'Map the Flow: Revealing Hidden Pathways of Information in VideoLLMs'. The dataset was created by author 'byminji' and was last updated on March 3, 2026.

Use Cases

• Instruction tuning of VideoLLMs based on video-text instruction pairs.
• Benchmarking model performance on video-based question answering and reasoning tasks.
• Analyzing information flow and hidden pathways within multimodal AI architectures.

Strengths

• Dataset is associated with a peer-reviewed ICLR 2026 conference paper.
• It is a cleaned version of an existing dataset, which suggests efforts to improve data quality.

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

Collection Method

Cleaned version of the VideoChat2-IT dataset.

Freshness

Last updated 2026-03-03 05:47:01; freshness should be verified.