Researchers Develop Audio-LLM Method to Filter Noisy Speech-to-Speech Translation Data
Researchers have developed a two-stage 'Rank-to-Distill' framework that trains an audio large language model to automatically identify and remove low-quality data from speech-to-speech translation corpora. Large-scale mined speech datasets commonly suffer from noise, misalignment, and semantic errors that degrade translation model performance. The method yields up to +1.4 ASR-BLEU improvement on benchmark datasets, offering a scalable alternative to manual data curation.
A study accepted to INTERSPEECH 2026 introduces a scalable pipeline for filtering training data used in end-to-end speech-to-speech translation (S2ST) systems. The core challenge is that large mined corpora, while abundant, frequently contain noisy or misaligned speech pairs that harm model quality. The proposed Rank-to-Distill strategy operates in two stages: a lightweight ranker first generates keep/drop pseudo-labels from noisy speech pairs without requiring manual annotation, and those labels are then used to train an audio large language model (Audio-LLM) to make filtering decisions directly from raw paired audio. This joint modeling approach allows the system to assess both acoustic fidelity and cross-lingual semantic consistency simultaneously. Experiments conducted on the CVSS-C and SpeechMatrix benchmarks demonstrated consistent improvements over unfiltered baselines, with gains of up to +1.4 ASR-BLEU. The work highlights the potential of Audio-LLMs as data quality gatekeepers in speech translation pipelines, reducing reliance on costly human labeling.
What's missing
The paper does not report computational cost or latency of the Rank-to-Distill pipeline relative to simpler filtering baselines, nor does it evaluate generalization to language pairs beyond those in CVSS-C and SpeechMatrix. It is also unclear how the method performs when the ranker's pseudo-labels are themselves highly noisy, which is a potential limitation in low-resource settings.
What different sources said
- arXiv cs.CLCenter
Leveraging Audio-LLMs to Filter Speech-to-Speech Training Data
Related
Gut Bacteria Enzyme Found to Break Down Heat-Processed Food Compounds, Producing Novel Biogenic Amines
Researchers have discovered that an enzyme in common gut bacteria can degrade N-epsilon-carboxymethyllysine (CML), a compound formed during thermal food processing, producing previously unknown biogenic amines. The enzyme, ornithine decarboxylase SpeC from enterobacteria, acts on CML and related modified lysine derivatives through a low-level 'underground' catalytic activity. This finding suggests a previously unrecognized communication axis between thermally processed dietary compounds and gut microbial physiology, with potential implications for host health.
Full-Length Gene Sequencing Reveals Two Distinct Bacterial Communities in Black-Legged Ticks Expanding Into Canada
Researchers used Oxford Nanopore full-length 16S rRNA gene sequencing to characterize the microbiome of Ixodes scapularis black-legged ticks collected in Nova Scotia, Canada, distinguishing between tick-adapted bacteria and environmentally acquired bacteria. The study comes as I. scapularis — the primary vector of Lyme disease — is rapidly expanding northward into Canada due to climate change. The findings suggest that environmentally derived bacteria in tick microbiomes are not mere contamination, which has implications for how tick microbiome data is collected and interpreted across surveillance studies.
Study Identifies Metabolic Link Between Cell Envelope Stress and Biofilm Formation in Bacteria
Researchers have discovered that the metabolite acetyl-CoA directly inhibits enzymes that degrade the bacterial signaling molecule c-di-GMP, connecting cell envelope biosynthesis stress to biofilm formation in Pseudomonas aeruginosa. The study found that sub-inhibitory concentrations of antibiotics targeting early peptidoglycan biosynthesis — but not other antibiotic classes — elevate c-di-GMP levels by reducing phosphodiesterase activity, with acetyl-CoA competing for the enzyme active site. Because the relevant enzyme domain is broadly conserved across bacterial species, this checkpoint mechanism may be widespread and could have implications for understanding antibiotic-induced biofilm responses.