Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
LoopGain, an open-source cost controller for AI agent loops, replaces fixed max_iterations with a control-theoretic stop-and-rollback policy, achieving 92.8% less API spend and ~15× faster median wall-clock time across 2…
Only 0.3% of entry-level data job listings mention AI skills, compared to 3.6% of senior roles, according to the AI Requirements Index from Datamata Studios, which tracks generative-AI and classic-ML skill mentions daily…
Scikit-ollama, a library based on scikit-llm, bridges scikit-learn with locally running Ollama models to perform zero-shot text classification without a cloud API. The library enables users to load a movie review sentime…
Hemispheric, a startup co-founded by Apple FaceID co-inventor Gidi Littwin, has raised $52 million to build a frontier AI model that decodes brain electrical activity to diagnose cognitive disorders without invasive proc…
A new benchmark, GLNBench, reveals that graph neural networks (GNNs) collapse under label noise when using strong backbones, with geometry-based methods like ERASE and UnionNET suffering rank collapse to near one on 512-…
A large-scale international study analyzing data from nearly 500,000 UK Biobank participants found that sleeping less than six hours or more than eight hours per night is associated with faster biological aging across mu…
A developer known as 'Dessinateur Projeteur' built a local-first engine called NexaVerify that runs code through 8 different AI models to catch hallucinations and logic errors in pull requests. The tool surfaces disagree…
Object detection models have evolved from two-stage detectors like R-CNN to one-stage models such as YOLO and transformer-based architectures like DETR, each offering trade-offs between accuracy and speed. Anchor-free an…
UKG VP of Engineering Thomas Newton described how his team built a Manager AI Adoption Dashboard that combines AI usage patterns, delivery outcomes measured by DX's TrueThroughput metric, and spend data to help engineeri…
An analysis of submissions to the NeurIPS Mechanistic Interpretability Workshop reveals a rise in AI-generated content, or "slop," since 2024, according to workshop organizer @andyarditi. The study, based on data from th…
AI is transforming supply chains by enabling compliance with regulations and adaptation to shocks, according to the Financial Times Technology section. The technology helps companies predict disruptions, optimize invento…
Geoffrey Hinton, the British-Canadian computer scientist whose foundational work on neural networks and deep learning underpins modern artificial intelligence, won the 2024 Nobel Prize in Physics jointly with John Hopfie…
RESOURCE2SKILL, a new framework from an unnamed research team, transforms tutorial videos and multimodal content into executable skills for AI software agents, boosting agent performance by an average of 11.9 percentage …
A developer ported Karpathy's nanochat to run on a TPU v6e-8 using JAX, achieving a CORE score that reproduces the original quality but with a model FLOPs utilization (MFU) of about 24%, roughly half of the 47-48% MFU Ka…
The Neuro-Symbolic Knowledge- and Gradient-Guided Reinforcement Learning (KGRL) algorithm, developed to exploit domain knowledge stored in a Datalog knowledge base, significantly outperforms state-of-the-art RL baselines…
Researchers have introduced Causal Rules and Maximally Specific Causal Relationships (MSCRs) to resolve statistical ambiguity in AI, building on Nancy Cartwright's definition of causes. The approach refines causal rules …
A new study reveals that intermediate representations from models like CLIP and DINOv2, combined with semantically aligned references, significantly improve synthetic image attribution accuracy. The research shows that a…
Researchers enhanced Ensemble Determinization Monte Carlo Tree Search (MCTS) with dynamic resource allocation strategies, including Dynamic Number of Determinizations and Dynamic Simulation Allocation, achieving statisti…
Google researchers introduced DiffusionGemma, a 26-billion parameter mixture-of-experts speech recognition model that uses discrete diffusion to refine whole transcripts in parallel, achieving a 6.6% word error rate on L…
A novel self-evolving critic framework improves LLM agent performance by learning from past actions, reducing Expected Calibration Error (ECE) by up to 54% across three agent benchmarks and three critic backbones without…