DialSeg-Ar - OPUS News: leaderboard
Metric: WindowDiff (0-1, lower is better; share of sliding windows over the utterance sequence whose predicted boundary count differs from the gold count; test split of OPUS News Commentary in Modern Standard Arabic; LLMs prompted zero-shot, DialSeg-Ar-Gemma3-4B fine-tuned on the training split). Source: arxiv.org. Saturation forecast: Estimated already saturated. 5 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Fanar-1-9B-Instruct | 0.57 |
| 2 | ALLaM-7B-Instruct-preview | 0.65 |
| 3 | Gemma 3 4B (IT) | 0.75 |
Interactive version: theaggregate.ai/benchmark?slug=dialseg-ar-opus-news · How It Works · Data refreshed daily, snapshot 2026-09-26.