Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers have developed a method to compress key-value caches in AI models using contractive iterated-map codes, achieving 36-54x memory reduction on GPT-2 with a perplexity cost of 11-15%. The technique also enables …
Researchers introduced CompDiff, a generative model that improves fairness in medical imaging by addressing demographic imbalances through a Hierarchical Conditioner Network. In tests on chest X-rays and fundus images, C…
A developer built a centralized vector database service for multi-project Retrieval-Augmented Generation (RAG) systems, allowing each project to have isolated storage while sharing one embedding model and API. The servic…
A developer explains why the normal distribution is fundamental in machine learning and statistics, highlighting its role in the Central Limit Theorem, statistical methods, and algorithm performance. The post covers key …
A developer created a black-box test plan for MonkeyCode, an AI task management system, to verify reliability under failures like client disconnects, worker crashes, and model timeouts. The plan defines invariants and fi…
The European Central Bank has upgraded its economic forecasting models with AI and machine learning to better predict energy price shocks, finding that inflation will not return to its 2% target until 2028. The recalibra…
Google Research, Google DeepMind, and university collaborators introduced SensorFM, a wearable health foundation model pretrained on over one trillion minutes of unlabeled sensor data from 5 million participants. The mod…
Google Research introduced SensorFM on July 9, 2026, a foundation model pre-trained on over one trillion minutes of de-identified wearable sensor data from five million participants. The model, which learns from Fitbit a…
Researchers warn that human-elephant conflicts in southern Africa are intensifying due to human population growth, land-use expansion, and climate-driven water deficits, with machine learning models predicting a 33-100% …
Researchers have developed a new structured pruning method for large language models that borrows from unstructured Adaptive Feature Retention, addressing distribution mismatches, sign information loss, and outlier influ…
Researchers developed a new method for out-of-scope intent detection using MiniLM embeddings, treating it as a one-class classification task with multi-cluster boundary learning. The approach outperformed existing baseli…
DeepSearch-Evolve introduces a self-distillation framework for training web agents in a controlled environment, achieving state-of-the-art results on benchmarks like BrowseComp, GAIA, and HotpotQA without distillation fr…
Researchers have introduced a new causal inference framework called triangulation that combines estimates from multiple models with different assumptions to infer true causal effects, avoiding the need to commit to a sin…
PhasorFlow, an open-source Python library, introduces a lightweight computing paradigm based on the S^1 unit circle for AI tasks, offering efficient alternatives to traditional algorithms. Its innovations include the Var…
A developer published a visual and practical guide to large language models, explaining that they are computer programs trained on massive text datasets to predict the next word. The guide covers how models like GPT-4 wo…
A developer at kuryzhev.cloud details common failure modes when deploying predictive alerting models on Prometheus data, where offline precision of 94% drops due to mismatches between training and live inference queries.…
Researchers applied a Spectral Temporal Graph Neural Network (StemGNN) to predict user equipment scheduling states in 5G networks, recovering 57-73% of sum rate loss caused by backhaul latency. The model achieved 87.57% …
Researchers developed a Q-learning-based adaptive retraining method to address traffic drift in Open Radio Access Networks (O-RAN), reducing retraining overhead while maintaining forecasting accuracy. The approach, which…
Researchers have developed NamedCurves+, an image editing tool that combines color naming and deep learning to enable intuitive, customizable enhancements. The system uses tone curves and transformer-based spatial depend…
Researchers introduced the PS4 framework, a proxy-supervised training method for target speaker extraction in conversational AI, using a bilingual dataset of 71,771 samples. The framework achieved second place on the REA…