inversedMixup: New Data Augmentation Method Combines Mixup Controllability with LLM Interpretability
Researchers have proposed inversedMixup, a data augmentation framework that merges the controllability of Mixup with the human-interpretability of large language model (LLM)-based generation by inverting mixed embeddings back into readable text. The method aligns a task-specific model's output embedding space with an LLM's input embedding space, enabling mixed latent representations to be reconstructed as natural language sentences at a controllable mixing ratio. This approach provides the first empirical evidence of the manifold intrusion phenomenon in text Mixup and demonstrates effectiveness in both few-shot and fully supervised settings.
inversedMixup is a unified data augmentation framework introduced by Fanshuang Kong and colleagues, submitted to arXiv in January 2026 and revised through June 2026. Traditional Mixup augmentation interpolates inputs and labels in latent embedding space but produces outputs that are not human-readable, while LLM-based augmentation generates interpretable text but offers limited control over the generation process. inversedMixup bridges this gap by leveraging recent advances in LLM inversion — the reconstruction of natural language from embeddings — to convert mixed embeddings into readable sentences with a controllable mixing ratio. A key contribution is the first empirical demonstration of manifold intrusion in text Mixup, a phenomenon where interpolated samples fall outside the natural data manifold, which the authors address through a dedicated mitigation strategy integrated into a three-stage augmentation pipeline. Extensive experiments reported in the paper show the method generalizes well across both few-shot and fully supervised classification scenarios.
What's missing
Open questions include how inversedMixup scales to very large LLMs, whether the manifold intrusion mitigation strategy generalizes beyond text classification, and how sensitive results are to the choice of inversion model.
What different sources said
- arXiv cs.CLCenter
inversedMixup: Data Augmentation via Inverting Mixed Embeddings
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.