KMP-Bench (Dialogue) - Challenge: leaderboard
Metric: Challenge principle accuracy (%): win rate averaged over the three challenge-specific criteria on the turns that target this principle, on KMP-Dialogue: the tutor model continues truncated multi-turn K-8 math tutoring dialogues under a dialogue-specific system prompt naming the target pedagogical principle, and a Gemini-2.0-Flash judge compares each response with the original tutor turn (Win/Tie/Lose); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude 3.7 Sonnet | 80 | #241 |
| 2 | Llama 3.3 70B Instruct | 59.2 | #520 |
| 3 | Mistral Nemo Instruct (2407) | 53.5 | #916 |
| 4 | Qwen 2.5 32B Instruct | 52 | #491 |
| 5 | Gemini 2.0 Flash | 51.1 | #331 |
| 6 | Qwen 2.5 72B Instruct | 50.5 | #436 |
| 7 | Qwen 2.5 14B Instruct | 48.7 | #634 |
| 8 | GPT-4o | 43.4 | #333 |
| 9 | Phi-4 | 41.3 | #701 |
| 10 | Qwen 2.5 7B Instruct | 41.2 | #846 |
| 11 | Llama 3.1 8B Instruct | 33 | #1018 |
| 12 | Qwen2.5-Math-72B-Instruct | 25.6 | #708 |
| 13 | Qwen2.5-Math-7B-Instruct | 17.5 | #1422 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kmp-bench-dialogue-challenge · How It Works · Data refreshed daily, snapshot 2026-10-11.