STAR: New Routing Method Improves Mixture-of-Experts Model Efficiency
Researchers have proposed STAR (Structure-Aware Routing), a new routing mechanism for Mixture-of-Experts (MoE) neural networks that incorporates principal subspace learning to better align routing decisions with input structure. Standard MoE routing relies on shallow linear projections that often fail to capture meaningful input patterns, leading to unstable expert assignments. The work, accepted at ICML 2026, addresses a core bottleneck in scaling large AI models efficiently.
Mixture-of-Experts (MoE) architectures scale model capacity by routing each input to a specialized subset of expert modules, but the effectiveness of this approach depends heavily on the quality of the routing mechanism. Current routing methods typically use simple linear projections that lack awareness of the underlying structure of inputs, which can result in unstable or poorly specialized routing. STAR addresses this by framing routing as a subspace learning problem, augmenting standard learnable routing with an evolving principal subspace estimated via the Generalized Hebbian Algorithm (GHA), which continuously tracks dominant structural patterns in the input. By aligning routing decisions with this learned subspace, STAR promotes more stable and meaningful expert specialization. The method was evaluated on synthetic benchmarks as well as large-scale language and vision tasks, consistently outperforming strong MoE baselines. An optional test-time subspace update mechanism further improves robustness when the input distribution shifts at inference time. The paper has been accepted at the International Conference on Machine Learning (ICML) 2026.
What's missing
The abstract does not specify the computational overhead introduced by maintaining and updating the evolving principal subspace via GHA, nor does it detail the scale (parameter count, dataset size) of the large-scale language and vision experiments used for evaluation. The degree of improvement over baselines in absolute terms is also not reported.
What different sources said
- arXiv cs.AICenter
STAR: Rethinking MoE Routing as Structure-Aware Subspace Learning
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.