Chinese Competitive Debating - Best Debater: leaderboard
Best-debater prediction in a benchmark built from 148 real Chinese-language debate matches, each adjudicated by three professional judges under a shared rubric (Yang et al., 2026). From the full match transcript the model distributes nine best-debater votes as the judges do; accuracy in percent, counting a match correct when its top pick is among the debaters with the most judge votes, higher is better.
Metric: Best-debater accuracy (%; the model distributes nine best-debater votes among the match's debaters from the full transcript; correct when its top debater is among those with the most judge votes; 148 competitive Chinese-language debate matches (4v4 and 2v2 formats), each adjudicated by three professional judges under a shared rubric; zero-shot on manually verified transcripts with fixed prompts; random choice 18.3). Source: arxiv.org. Saturation forecast: Around May 2028. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Medium) | 56.8 | #29 (GPT-5.6 Sol) |
| 2 | DeepSeek V4 Pro | 55.4 | #181 |
| 3 | Claude Opus 5 (Medium) | 52 | #20 (Claude Opus 5) |
| 4 | DeepSeek V4 Flash (Reasoning) | 43.9 | #254 (DeepSeek V4 Flash) |
| 5 | DeepSeek V4 Flash | 43.2 | #254 |
| 6 | GPT-5.6 Luna (Medium) | 43.2 | #117 (GPT-5.6 Luna) |
| 7 | Claude Sonnet 5 (Medium) | 41.9 | #108 (Claude Sonnet 5) |
| 8 | Gemini 3.5 Flash Lite (Medium) | 34.5 | #300 (Gemini 3.5 Flash Lite) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=chinese-competitive-debating-best-debater · How It Works · Data refreshed daily, snapshot 2026-11-12.