Machine Learning News

A data-driven control framework combining surrogate modeling and reinforcement learning achieved an almost 2x improvement over operators' actions in optimizing target polarization in nuclear physics scattering experiment…

A Mixture-of-Agents (MoA) study measuring per-token sufficient compute found that a 0.5B agent reproduces 92–95% of reference tokens on three core benchmarks, while the most expensive 10% of tokens account for 64–80% of …

A hybrid Raman spectroscopy framework called HyMLRaman identified six pharmaceutical compounds with 96.31% accuracy and a 96.36% macro-F1 score using an EfficientNet-B3 plus SVM configuration, according to an arXiv paper…

Researchers introduced HXAI, a hierarchical framework that combines explainable AI with differential privacy to give grid operators zonal-level insights into household energy consumption without exposing appliance-level …

A study using gradient boosting and Mondrian Conformal Prediction, posted to arXiv as 2610.02225v1, found that the gap between companies' self-reported emissions and an algorithmic baseline built from U.S. SEC financial …

A September 30, 2026 arXiv report by Ling Min Serena Khoo systematically compares several nearest-neighbour retrieval variants for molecular fingerprint prediction from MS/MS spectra, showing how differing assumptions ab…

Researchers introduced Rank-Aware Speculative Sampling (RASS), a verification rule for speculative draft trees based on rank-aware list coupling, which improves on Diffusion Greedy Rejection Sampling (D-GRS) across evalu…

Researchers introduced ReRoute, a framework for targeted scientific what-if prediction that combines factual data with partial mechanistic knowledge and requires no controlled intervention data for adaptation, according …

A new arXiv paper (2610.02253v1) establishes approximation rates for dropout ReLU networks, showing that the unit ball of W^{n,∞}([0,1]^d) can be approximated uniformly by constant-depth networks of size Õ_{n,d}(p^{-9}ε^…

A study posted to arXiv (2610.02284v1) found limited support for the hypothesis that natural interpulse-interval (IPI) variation contributes substantially to automated bat-species classification, with transformer model P…

A new arXiv paper (2610.02344v1) develops an early-training stability theory for Joint-Embedding Predictive Architectures (JEPAs) that explains representation collapse through a per-mode stability ratio μ_i = γ_i / σ_i, …

Princeton researchers introduced SCION, a hierarchical compositional scene representation that replaces millions of independent 3D Gaussians with a compact vocabulary of reusable primitives and lightweight world-space in…

FactorSplat, a per-scene N-dimensional Gaussian splatting proxy from researchers behind the arXiv paper 2610.02382v1, accepts region-specific intensity-to-RGBA transfer function curves at inference, letting one checkpoin…

A multi-stage AI framework for assessing androgenetic alopecia (AGA) from low-magnification scalp images achieved a test mAP@0.5 of 0.920 and recall of 0.860 with YOLOv8m under patient-disjoint evaluation, according to a…

A new arXiv paper, arXiv:2610.02494v1, proposes DeepStratNet, a context-aware coordinate regression framework that reformulates 3D seismic horizon tracking as a bounded coordinate regression problem instead of dense sema…

Researchers introduced IG-VLA, a vision-language-action reasoning framework that imagines task-relevant future scene evolution in visual representation space and internalizes it into a compact Scene Gist Token, according…

NVIDIA released Physical AI Smart Spaces, a benchmark it calls the first to simultaneously provide large-scale, multi-class, multi-camera 3D perception data for indoor smart spaces, containing over 280 hours of synchroni…

Training Qwen3-4B with counterexample-only supervised fine-tuning collapsed true-theorem recognition from 0.27 to 0.00, while reinforcement learning with a sparse outcome-only reward repaired the collapse and exceeded th…

A new arXiv paper (2610.02460v1) introduces Calibrated User Embeddings (CUE), a framework that encodes observed user sessions and samples continuous representations decoded into persona commands to steer LLMs as user sim…

A systematic study of prompted large language models in time and event expression extraction, published as arXiv:2610.02549v1, found that strong base-task performance generally predicts better generalization but that thi…