AMI-ME - Topic Segmentation: leaderboard
Topic segmentation of AMI Corpus meetings on AMI-ME (2026), which divides each meeting into fine-grained topic segments for effectiveness scoring. WindowDiff error from 0 to 1 against the reference segmentation; lower is better.
Metric: WindowDiff segmentation error (0-1) of the model's topic boundaries against the AMI-ME reference fine-grained segmentation (130 AMI Corpus meetings); lower is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | DeepSeek R1 Distill Llama 70B | 0.3658 | #1676 | |
| 2 | Qwen 3 32B | 0.3713 | #983 | |
| 3 | GPT-4o | 0.398 | #754 | |
| 4 | Llama 3.3 70B Instruct | 0.4108 | #1394 | |
| 5 | Gemini 2.5 Flash | Non-reasoning | 0.4486 | #466 |
No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=ami-me-topic-segmentation · How It Works · Data refreshed daily, snapshot 2026-11-22.