LongVT-Parquet: Training Annotations and Evaluation Benchmark for Long Video Reasoning

LongVT-Parquet provides the training data annotations and evaluation benchmark for the LongVT project. The dataset supports an end-to-end agentic framework for 'Thinking with Long Videos' via interleaved Multimodal Chain-of-Tool-Thought. It was created by 'longvideotool' and last updated on March 9, 2026.

Use Cases

• Train multimodal reasoning models based on the described interleaved Chain-of-Tool-Thought framework.
• Benchmark agentic video question-answering systems based on the provided evaluation data.
• Develop tool-calling capabilities for long video analysis based on the multimodal agentic framework.

Strengths

• Dataset is explicitly designed as a benchmark for evaluating long video reasoning systems.
• Data is formatted in Parquet, a columnar storage format efficient for large-scale data processing.
• Last updated on March 9, 2026, 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.
• The description notes privacy-related modifications to media paths, requiring users to consult the full description page.

Provenance

Source

huggingface

Collection Method

Created as part of the LongVT project; specific collection method is not detailed.

Time Range

null

Freshness

Last updated 2026-03-09 06:46:41.

Geography

null