HINTT Submission to the 2nd MLC-SLM Challenge: Comparing Cascaded and Unified Approaches to Diarization and ASR

Key points

• We address multilingual speaker-attributed ASR, where systems must determine who spoke when and what was spoken.1
• We investigate two modeling strategies for this problem: a cascaded pipeline that combines speaker diarization with speech-LLM-based ASR, and a unified speech LLM that directly generates speaker labels, timestamps, and transcriptions.1
• Our final submission is based on the cascaded pipeline, consisting of a fine-tuned DiariZen diarization model, a fine-tuned Qwen3-ASR model, and LLM-based generative error correction.1
• Experimental results demonstrate that the cascaded system remains more reliable under the MLC-SLM Task 1 conditions, while unified speech LLMs offer a promising direction for future speaker-attributed ASR.1

Sources (1)

• [1]HINTT Submission to the 2nd MLC-SLM Challenge: Comparing Cascaded and Unified Approaches to Diarization and ASR
  arXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 6, 09:57 AM

  This paper presents the HINTT system submitted to the 2nd Challenge and Workshop on Multilingual Conversational Speech Language Model (MLC-SLM).

  We address multilingual speaker-attributed ASR, where systems must determine who spoke when and what was spoken.

Extractive summary: sentences quoted from the sources.