RESCUE-BENCH - Relation Pattern Prediction: leaderboard
RESCUE-BENCH: identifying which relation pattern holds between the speakers, since the right support depends on it. Accuracy, higher is better. Source: arxiv.org/abs/2609.09657.
Metric: Accuracy (%). Source: arxiv.org. Saturation forecast: Around July 2028. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 45.6 | #112 |
| 2 | Kimi K2.5 | 45.26 | #153 |
| 3 | Qwen 3 Max | 43.22 | #213 |
| 4 | Qwen 3.5 Plus | 40.27 | #136 |
| 5 | MiniMax-M2.5 | 39.85 | #316 |
| 6 | DeepSeek V4 Flash | 39.39 | #155 |
| 7 | GPT-4o | 38.69 | #357 |
| 8 | Qwen Plus | 38.27 | #393 |
| 9 | DeepSeek R1 | 37.89 | #265 |
| 10 | DeepSeek V3.2 | 36.08 | #211 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #6 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rescue-bench-relation-pattern-prediction · How It Works · Data refreshed daily, snapshot 2026-11-06.