RESCUE-BENCH - Support Target Prediction: leaderboard
RESCUE-BENCH: choosing who in a multi-party conversation should receive the support. MRR, higher is better. Source: arxiv.org/abs/2609.09657.
Metric: MRR (%). Source: arxiv.org. Saturation forecast: Around February 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V4 Pro | 82.23 | #112 |
| 2 | Qwen 3.5 Plus | 81.57 | #133 |
| 3 | GPT-4o | 80 | #357 |
| 4 | Qwen Plus | 79.21 | #393 |
| 5 | DeepSeek V4 Flash | 77.61 | #155 |
| 6 | DeepSeek V3.2 | 77.14 | #210 |
| 7 | MiniMax-M2.5 | 77.01 | #316 |
| 8 | DeepSeek R1 | 76.88 | #262 |
| 9 | Qwen 3 Max | 76.28 | #213 |
| 10 | Kimi K2.5 | 75.6 | #152 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=rescue-bench-support-target-prediction · How It Works · Data refreshed daily, snapshot 2026-11-05.