# Pushing the Limits of End-to-End Diarization

arXiv: https://arxiv.org/abs/2509.14737 · HTML: https://mlsanity.com/paper/2509.14737/

## Key facts

- Cited by 2 papers in the map (data through 19 Oct 2026).
- First posted on arXiv 18 Sep 2025.

## Built on by (most cited first, 2 of 2)

- [Speech Language Models for Full-Meeting Speaker Diarization: Capabilities and Limitations](https://mlsanity.com/paper/2609.23114/) (2026)
- [Beyond DER: Speaker Counting in Crowded End-to-End Diarization](https://mlsanity.com/paper/2609.27315/) (2026)

## Abstract (arXiv)

In this paper, we present state-of-the-art diarization error rates (DERs) on multiple publicly available datasets, including AliMeeting-far, AliMeeting-near, AMI-Mix, AMI-SDM, DIHARD III, and MagicData RAMC. Leveraging EEND-TA, a single unified non-autoregressive model for end-to-end speaker diarization, we achieve new benchmark results, most notably a DER of 14.49% on DIHARD III. Our approach scales pretraining through 8-speaker simulation mixtures, ensuring each generated speaker mixture configuration is sufficiently represented. These experiments highlight that EEND-based architectures possess a greater capacity for learning than previously explored, surpassing many existing diarization solutions while maintaining efficient speeds during inference.