CourtSI-Bench - Player-Line: leaderboard
Metric: Exact-match accuracy (%) on the 495 player-to-court-line relation questions, CourtSI-Bench's 3,686 human-verified spatial questions on single broadcast frames of badminton, tennis and table tennis (players marked by boxes), answers read by exact format match from the model's raw output; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 25 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.2 | 77.4 | #105 |
| 2 | Seed 1.8 | 77.4 | #136 |
| 3 | Qwen 3 VL 235B A22B | 71.1 | #228 |
| 4 | InternVL3-8B | 56.6 | #606 |
| 5 | Qwen 2.5 VL 7B Instruct | 54.9 | #643 |
| 6 | Qwen 3 Max | 51.3 | #201 |
| 7 | Grok 4 | 48.9 | #169 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=courtsi-bench-player-line · How It Works · Data refreshed daily, snapshot 2026-10-11.