Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers propose SQS, a Bayesian method that unifies pruning and quantization for deep neural network compression by modeling weights with sparse quantized sub-distributions, achieving higher compression rates while p…
Researchers found that transformers can encode an inferred partner's expertise in their residual stream at an early layer, before that information influences the model's output, extending prior findings on directly state…
A newly disclosed training method injects deliberate weight perturbations into neural networks via a specialized regularizer, securing the first layer so protections propagate across all subsequent layers, according to a…
A patent-pending surgical tool combining on-device artificial intelligence tissue classification with haptic feedback has reached Technology Readiness Level 3 and is available for licensing. The device uses a laser-based…
An engineer replaced the message function in a graph attention network with a 4-qubit variational quantum circuit, creating QTGNN v2 for stock movement prediction. The model uses a dynamic graph of 10 S&P 500 stocks with…
A new benchmark and dataset called RoboSPA evaluates Vision-Language-Action (VLA) models on robotic manipulation tasks with increased spatial and procedural complexity, moving beyond simple scenes and short-horizon tasks…
James Wilkinson, a Turing Award winner in 1970, introduced backward error analysis, a method that measures how much input perturbation is needed to make an algorithm's output exactly correct. This approach became the gol…
A developer built Video Transcriber Pro, a desktop tool that transcribes video and audio locally using OpenAI's Whisper model, ensuring sensitive material never leaves the machine. The pipeline integrates yt-dlp, ffmpeg,…
Researchers introduced TransNormal-2, a geometry-grounded rectified flow model with edge-aware decoding that addresses VAE reconstruction degradation in monocular normal estimation, improving precision at object boundari…
A new arXiv preprint (submitted September 7, 2026) proposes that efficient exploration in reinforcement learning should prioritize generating generalizable experience, and shows theoretically that optimally efficient exp…
Researchers introduced Gander, an end-to-end model unifying omni perception, realtime interaction, and agentic capabilities in a single framework, capable of continuously processing streaming multimodal inputs such as vi…
Researchers propose Mask Forcing, a dual-noise masking rollout technique that improves autoregressive video diffusion distillation by reducing over-saturation artifacts in generated videos. The method targets AR video di…
Researchers introduced Kalman Delta Networks (KDN), a linear-attention architecture that models associative memory as an online Bayesian filtering problem, using a Kalman filter to update memory with uncertainty-aware ga…
Researchers propose a novel framework that enriches feedback by using environments as scaffolds to bootstrap self-evolving agents in long-horizon tasks, addressing reward sparsity in reinforcement learning for large lang…
Princeton University's PACMAN AI system can control fusion plasma in milliseconds and predict dangerous instabilities before they occur, addressing disturbances that develop within thousandths of a second in plasma hotte…
An engineer detailed the importance of automation in MLOps for scaling AI deployment pipelines, emphasizing model registries, versioning, and CI/CD principles adapted for machine learning. The article outlines a structur…
Researchers introduced NeoHorse-1, a family of agent-native models designed for recursive self-improvement through agentic post-training with a routing harness. The system observes its own capabilities and converts that …
Zepto, an Indian quick-commerce platform, reports that its evaluation-first AI agents, built on Databricks and MLflow, now handle over 80% of customer support tickets, cutting support costs by 65% and achieving payback i…
At a recent event, speakers including Aritra Roy Gosthipaty from Hugging Face emphasized the need for India to focus on machine learning infrastructure builders rather than API-level developers, highlighting practical te…
Researchers introduced Continual Harness, a reset-free self-improving system for embodied agents that automates online adaptation without human intervention, and reported that their Gemini Plays Pokemon (GPP) experiments…