AI Briefing#machine-learning
The latest AI and developer news about #machine-learning, with the original source and a short summary.
The latest AI and developer news about #machine-learning, with the original source and a short summary.
The new version features a 4x speedup in neighbour search and a recursive initialization method that improves cluster resolution on million-scale datasets.
The system uses a Resource Core Model with 6B sparse parameters to predict marginal utility, directing compute to high-impact requests across a heterogeneous fleet.
Mitra-v2, a second-generation tabular foundation model, is pretrained on tables with 10x more rows and 3x more columns than its predecessor, leading TabArena and ScoringBench.
MUSINSA used its campaign meta-engine to automatically optimize over half of the hundreds of millions of messages sent monthly.
The benchmark uses 5,800 labeled combinations where every test peptide is at least four edits away from any publicly reported TCR specificity.
The system preserves F3's shared computation efficiency while allowing independent feature validation and rollback.
The free tool extracts results tables and ablations from 79k papers across ICML, NeurIPS, ICLR, CVPR, and AAAI since 2016 to track benchmark records and baselines.
MUSINSA has developed the DeciML and NeXA platforms to handle decision-making and discovery beyond prediction.
A new theorem demonstrates that Joint-Embedding Predictive Architectures (JEPAs) achieve cross-modal alignment with 42 times fewer training samples than next-token prediction.
The new architecture uses a Mixture-of-Experts model and fused GPU kernels to deliver near-ranking-level performance at 300x fewer FLOPs than the 2024 system.
The new operator uses exponential relevance density over a continuous memory curve to achieve exact normalized attention with closed-form gradients.
The ensemble achieves a MASE of 0.683 on GIFT-Eval, outperforming its best individual member without requiring fine-tuning.
The pre-trained foundation model requires no feature engineering or table flattening, using up to 10,000 historical examples for single-pass inference on databases with over 500 billion rows.
The new system generates global 3 km climate data at approximately 1/300th the cost of the X-SHiELD atmospheric model.
The open-source library supports non-commutative geometric spaces like hyperbolic and SO(n) rotations to preserve event order in sequence modeling.
The method achieves a mean F1 score of 0.798 across 14 language pairs in CS-FLEURS using the Whisper-small encoder.
The new pipeline reduces median post-push scan execution time by 86% and improves user acceptance by 11.5% compared to the previous LLM approach.
Google teams have published 90 peer-reviewed papers with over 90,000 citations and provided over $60 million in funding to health research organizations in 2026.
Meta aggregates fragmented idle inference capacity into a shared pool to enhance ad relevance without requiring new hardware.
Apple researchers introduced Normalizing Trajectory Models (NTM), a framework that matches strong baselines in four sampling steps while retaining exact likelihood over the generative trajectory.
A study involving 2,000 mothers showed AI can interpret blind-sweep ultrasounds as accurately as trained sonographers.
The study identifies four key sites where users negotiate ontological boundaries when defining their own phenomena for machine learning training.
The paper introduces feature-splitting and physical sparsity priors to enable hierarchical DSR models to predict bifurcations and unseen dynamical regimes without explicit control parameter knowledge.
The system provides 30 to 60 minutes of advance warning and detected nearly 80% of major space-weather events during evaluation.