DialSeg-Ar - Code-Switched Podcasts: 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 Gulf Arabic and English code-switched podcast transcripts; 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 | Gemma 3 4B (IT) | 0.67 |
| 2 | Fanar-1-9B-Instruct | 0.7 |
| 3 | ALLaM-7B-Instruct-preview | 0.81 |
Interactive version: theaggregate.ai/benchmark?slug=dialseg-ar-code-switched-podcasts · How It Works · Data refreshed daily, snapshot 2026-09-26.