VideoTemp-o3: Video Question-Answer Pairs for Agentic Temporal Grounding

VideoTemp-o3 is a dataset for harmonizing temporal grounding and video understanding in agentic thinking-with-videos. The dataset contains question and answer pairs sourced for training the VideoTemp-o3 model, which performs on-demand temporal grounding to locate relevant video segments. It was created by Kwai-Keye and last updated on May 20, 2026.

Use Cases

• Training models for video question-answering based on described agentic pipeline.
• Benchmarking temporal grounding algorithms based on iterative segment refinement.
• Developing multimodal agents that reason over video evidence.

Strengths

• Dataset is designed for a specific, advanced AI task combining temporal grounding and video QA.
• The agentic pipeline is described, suggesting a structured approach to data use.

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

Kwai-Keye

Collection Method

Question and answer pairs sourced for training the VideoTemp-o3 model.

Freshness

Last updated 2026-05-20 12:39:41; freshness should be verified.