Medvidu Eccv2026 Trainval: Medical Video Understanding Benchmark Split

MedVidU_ECCV2026_TrainVal is a public training and validation split for the MedVidU Challenge at the ECCV 2026 Workshop on Medical Video Understanding. The split is derived from the benchmark introduced in the MedGRPO paper presented at CVPR 2026. It was authored by UII-AI and last updated on May 14, 2026.

Use Cases

• Training models for medical video understanding based on the described benchmark tasks.
• Local validation of model performance on heterogeneous medical video tasks.
• Developing multi-task reinforcement learning approaches for medical video data.

Strengths

• Split is derived from a benchmark introduced in a CVPR 2026 paper, suggesting a research foundation.
• Explicitly intended for public use by challenge participants.

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

UII-AI

Collection Method

Derived from the benchmark introduced in MedGRPO: Multi-Task Reinforcement Learning for Heterogeneous Medical Video Understanding (CVPR 2026).

Freshness

Last updated 2026-05-14 15:31:31; freshness should be verified.