# LLM-ForcedAligner: A Non-Autoregressive and Accurate LLM-Based Forced Aligner for Multilingual and Long-Form Speech

arXiv: https://arxiv.org/abs/2601.18220 · HTML: https://mlsanity.com/paper/2601.18220/

## Key facts

- Cited by 9 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 26 Jan 2026.

## Built on by (most cited first, 9 of 9)

- [Qwen3-ASR Technical Report](https://mlsanity.com/paper/2601.21337/) (2026)
- [SPEARBench: A Benchmark for Naturalness Evaluation in Streaming Speech-to-Speech Language Models](https://mlsanity.com/paper/2607.05365/) (2026)
- [Gradient-Based Speech-to-Text Alignment for Any ASR Model: From CTC to Speech LLMs](https://mlsanity.com/paper/2607.06831/) (2026)
- [X2Streaming-ASR: wait when uncertain, emit when ready for streaming ASR](https://mlsanity.com/paper/2609.08672/) (2026)
- [X2-NativeCursor: Native-Token Text Progress Tracking for Incremental-Text Streaming Codec TTS](https://mlsanity.com/paper/2609.09677/) (2026)
- [FA-Bench: A Benchmark for Phone- and Word-Level Timestamp Accuracy in Forced Alignment and ASR on Clean and Noisy Speech](https://mlsanity.com/paper/2609.32396/) (2026)
- [From Script to Drama: An Agentic Framework for Controllable Multi-Speaker Dialogue TTS](https://mlsanity.com/paper/2609.33362/) (2026)
- [FuseAlign: Forced Alignment in the Wild](https://mlsanity.com/paper/2609.33650/) (2026)
- [Where Words Begin and End](https://mlsanity.com/paper/2610.17782/) (2026)

## Abstract (arXiv)

Forced alignment (FA) predicts start and end timestamps for words or characters in speech, but existing methods are language-specific and prone to cumulative temporal shifts. The multilingual speech understanding and long-sequence processing abilities of speech large language models (SLLMs) make them promising for FA in multilingual, crosslingual, and long-form speech settings. However, directly applying the next-token prediction paradigm of SLLMs to FA results in hallucinations and slow inference. To bridge the gap, we propose LLM-ForcedAligner, reformulating FA as a slot-filling paradigm: timestamps are treated as discrete indices, and special timestamp tokens are inserted as slots into the transcript. Conditioned on the speech embeddings and the transcript with slots, the SLLM directly predicts the time indices at slots. During training, causal attention masking with non-shifted input and label sequences allows each slot to predict its own timestamp index based on itself and preceding context, with loss computed only at slot positions. Dynamic slot insertion enables FA at arbitrary positions. Moreover, non-autoregressive inference is supported, avoiding hallucinations and improving speed. Experiments across multilingual, crosslingual, and long-form speech scenarios show that LLM-ForcedAligner achieves a 69%~78% relative reduction in accumulated averaging shift compared with prior methods. Checkpoint and inference code are available at https://github.com/QwenLM/Qwen3-ASR.