Semi-structured Retrieval Benchmark (STaRK) Leaderboard

Refer to the STaRK paper for details on metrics, tasks and models.

Model types

Submit Your Results

Submit your results to be included in the leaderboard. Please ensure your submission meets all requirements. For questions, contact stark-qa@cs.stanford.edu. Detailed instructions can be referred at submission instructions.

Dataset*
Split*
Model Type*

Select the appropriate category for your model


Recent Submissions and Updates

2026-02-12 01:30:04
✅ Status update: Papr v2 has been approved

2026-01-08 22:01:49
📥 New submission: Papr v2 on test-0.1/mag

2025-12-24 21:02:41
📥 New submission: cde on human_generated_eval/mag

2025-02-12 19:49:39
✅ Status update: Paprv1 has been approved

2025-02-05 06:50:41
📥 New submission: Paprv1 on test-0.1/mag