Hy-MultiTurn - Object Localization: leaderboard
The object localization dimension of Hy-MultiTurn (Tencent Hunyuan, 2026): the importance-weighted checklist score in percent on 35 long Chinese dialogues that require identifying the intended objects from compound object and attribute conditions. Higher is better.
Metric: Importance-weighted score (%) on the 35 object localization dialogues, 209 long Chinese multi-turn dialogues (12 to 76 turns), three rolls per task averaged, three LLM judges (GPT-5.2, Qwen 3.6, DeepSeek-V4 Pro) averaged, a deterministic checker for precise execution; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 22 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Grok 4.5 (High) | 95.2 | #82 (Grok 4.5) |
| 2 | Kimi K3 (Max) | 92.1 | #54 (Kimi K3) |
| 3 | Claude Opus 4.6 (High) | 92 | #102 (Claude Opus 4.6) |
| 4 | GPT-5.5 (High) | 90.7 | #49 (GPT-5.5) |
| 5 | Claude Opus 4.7 (High) | 90.7 | #79 (Claude Opus 4.7) |
| 6 | GPT-5.4 (High) | 87.9 | #123 (GPT-5.4) |
| 7 | Gemini 3.1 Pro (Preview) (High) | 87.7 | #94 (Gemini 3.1 Pro (Preview)) |
| 8 | Hy3 (High) | 86.1 | #195 (Hy3) |
| 9 | Doubao-Seed-2.1-Pro (High) | 85.5 | |
| 10 | Kimi K2.6 | 82.3 | #178 |
| 11 | Kimi K2.5 (Thinking) | 78.2 | #256 (Kimi K2.5) |
| 12 | Hy3-preview (High) | 74.5 | #316 (Hy3-preview) |
| 13 | Qwen 3.6 Plus (Thinking) | 72.3 | #222 (Qwen 3.6 Plus) |
| 14 | Kimi K2.5 (Non-reasoning) | 60.6 | #256 (Kimi K2.5) |
| 15 | Kimi K2.6 (Non-reasoning) | 60.4 | #178 (Kimi K2.6) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=hy-multiturn-object-localization · How It Works · Data refreshed daily, snapshot 2026-11-13.