DecSelfMask: New Method Uses Unlabeled Text to Improve Medical Classification Tasks
Researchers have proposed DecSelfMask, a self-supervised training approach that uses relevance-guided masking to improve decoder-only language model performance on classification tasks with limited labeled data. The method was evaluated on 136 tasks drawn from 1.9 million clinical notes from an Italian hospital, testing five models of varying scales and architectures. It outperformed standard supervised fine-tuning by 19.9 Macro F1 points, offering a potentially significant advance for data-scarce domains like healthcare.
DecSelfMask (Decoder Self-learning by Masking) addresses a core challenge in applied NLP: classification tasks in specialized domains such as medicine often lack sufficient annotated training data. The approach extends conventional self-learning by using relevance attribution methods to identify which parts of unlabeled text are most task-relevant, then masking those portions to create self-supervised training examples. The model is trained to reconstruct the masked content via next-token prediction, with the hypothesis that this process encodes useful structural and semantic knowledge about the domain. Evaluated across 136 classification tasks on 1.9 million Italian clinical notes, DecSelfMask consistently outperformed competing approaches: +19.9 Macro F1 over standard supervised fine-tuning, +12.5 over synthetic label generation, and +6.3 over continual pretraining. The study also included a probing analysis to better understand what the models learn. The results suggest the method is robust across model families and scales, though it has so far only been validated in a single-language, single-institution clinical setting.
What's missing
The study is limited to Italian-language clinical notes from a single hospital, leaving open questions about generalizability to other languages, healthcare systems, and non-medical domains. The paper does not report computational cost comparisons between DecSelfMask and baseline methods, which is relevant for practical adoption. It is also unclear how sensitive the gains are to the choice of relevance attribution method used for masking.
What different sources said
- arXiv cs.CLCenter
DECSELFMASK: Leveraging Unlabeled Text via Self-Relevance-Guided Masking for Decoder-Only Classification
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.