ART: New Method Improves Efficiency of Large Language Model Decoding by Up to 20%
Researchers have introduced Item Response Scaling Laws (IRSL), a framework that dramatically reduces the computational cost of estimating how language model performance scales with size and compute. IRSL integrates Item Response Theory — a psychometric tool traditionally used in educational testing — into the scaling law estimation process, separating model ability from question difficulty to reduce parameter complexity from O(M×N) to O(M+N). The approach could make scaling law research far more accessible by requiring only 50 benchmark questions instead of tens of thousands.
A team of researchers has proposed Item Response Scaling Laws (IRSL), a unified framework designed to make the estimation of neural scaling laws significantly cheaper and more generalizable. Traditional scaling law derivation requires evaluating thousands of model checkpoints across millions of inference samples, making it prohibitively expensive. IRSL addresses this by incorporating Item Response Theory (IRT), a psychometric framework used in standardized testing, to disentangle a model's latent ability from the characteristics of individual benchmark questions. The framework is instantiated using Beta-IRT, which leverages continuous signals such as token probabilities and pass rates rather than binary correct/incorrect responses, capturing richer information. Validated across two scaling paradigms — pre-training downstream scaling with over 6,600 checkpoints and 37,000 questions, and test-time scaling with 12 models and 120 questions — IRSL achieves comparable or better decision accuracy using just 50 questions per benchmark after a one-time calibration. Additionally, the latent ability estimates learned by IRSL generalize across benchmarks that share the same measurement objective, enabling cross-benchmark performance forecasting.
What's missing
The paper does not yet report results from peer review or independent replication. Key open questions include how well IRSL generalizes to highly heterogeneous benchmark sets where the assumption of a shared measurement objective may not hold, how sensitive results are to the quality of the initial one-time calibration data, and whether the Beta-IRT parameterization introduces its own biases when model response distributions are highly skewed or near ceiling/floor performance levels.
What different sources said
- arXiv cs.LGCenter
Adaptive Generate-Rank-Verify: Inference-Time Search with Costly Verification
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.