Gradient-Based Speech-to-Text Alignment for Any ASR Model: From CTC to Speech LLMs
Quick Answer
This study introduces a gradient-based alignment method applicable to any differentiable ASR model, including speech LLMs, without requiring training or model modifications.
Quick Take
Evaluated across sixteen models, it shows promising alignment performance, particularly in scenarios where native alignments are weak, although it incurs a cost of one backward pass per token.
Key Points
- Gradient-based alignment works across all ASR models, including CTC and speech .
- No training or model modifications are required for this alignment method.
- Evaluated on TIMIT and Buckeye datasets, showing better performance in weak native alignments.
- Main drawback is the computational cost of one backward pass per token.
- Aligns on the input grid, improving temporal precision over encoder grid methods.
Paper Resources
📖 Reader Mode
~2 min readAbstract:Speech-to-text alignment means finding the temporal boundaries of each word in the audio. Some models provide such an alignment directly and others do not. Connectionist temporal classification (CTC) and transducer models have an alignment by construction, whereas attention-based encoder-decoders (AED) and speech large language models (LLMs) do not, and their word timings are usually read off the attention weights instead. All of these signals live on the encoder frame grid, which bounds their temporal precision. We study a generic gradient-based alignment that applies to any differentiable ASR model. We take the gradient of each teacher-forced token log probability with respect to the input, reduce it to a per-frame saliency, and decode the resulting matrix into word boundaries with a single dynamic-programming pass. The method needs no training, no model modification and no alignment heads, works across all model families including the speech LLMs, and aligns on the input grid rather than on the coarser encoder grid. We evaluate it on sixteen models from four families, on read (TIMIT) and spontaneous (Buckeye) speech, each against the model's own native or attention-based alignment. We find that the gradient yields a usable alignment for every model, that it is usually somewhat behind a strong native aligner but better where the native alignment is weak, as for the streaming models, and that its main disadvantage is the cost of one backward pass per token.
| Subjects: | Computation and Language (cs.CL); Artificial Intelligence (cs.AI); Computer Vision and Pattern Recognition (cs.CV); Machine Learning (cs.LG) |
| Cite as: | arXiv:2607.06831 [cs.CL] |
| (or arXiv:2607.06831v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2607.06831 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Albert Zeyer [view email]
[v1]
Tue, 7 Jul 2026 21:57:54 UTC (216 KB)
— Originally published at arxiv.org
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CL
See more →TriAgent: Divergence-Aware Committees for Cost-Efficient Financial Sentiment Analysis
TriAgent introduces a cost-efficient multi-agent system for financial sentiment analysis, combining VADER, FinBERT, and Qwen2.5. It achieves an F1 score of ~0.87 with significant savings of $9.3M/year at a 10M-user scale compared to GPT-4o-mini, while also detecting hallucinations with an AUC of 0.90.