Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers introduce token time continuous diffusion (TTCD), a new diffusion language model that operates in continuous space and incorporates per-token times, allowing some tokens to proceed from noise to token faster …
A study from arXiv:2607.14205 found that federated learning (FL) on radiology reports is vulnerable to privacy leakage via gradient inversion, with exact sentence reconstruction rates ranging from 31% to 44% across three…
Researchers propose LIGO-PINN, a framework using learned initialization via gated layerwise optimization to overcome convergence failures in physics-informed neural networks (PINNs). In evaluations on 1D and 2D PDE domai…
Researchers propose DCVC-Mamba (DCVC-MB), a neural video codec framework for B-frame coding that incorporates an IBP frame strategy, a spatio-temporal fusion model based on state-space models, and an entropy-aware skippi…
A new study from arXiv (2607.14246v1) finds that generative models have a fixed steering budget set by training data, beyond which knobs like prompts and guidance scales cannot move a property, but concrete examples can.…
Researchers introduce SD-MAR, a framework using synthetic data and reinforcement learning to train vision-language models on multi-image analytical reasoning tasks such as change detection and visual comparison. Fine-tun…
Preempt AI launched v2 of its AI security platform, offering ML-powered protection against prompt injection, jailbreaks, and data leaks with 99.65% accuracy trained on 589K+ real attack samples. The model-agnostic layer …
A developer's tutorial explains the complete architecture of vector databases, covering schema design, indexing methods like IVF and HNSW, product quantization, distance metrics, metadata indexing, and token economics to…
University of Illinois Urbana-Champaign researchers have released an open-source AI teaching assistant on HuggingFace that runs 11 separate models in parallel for text/image retrieval, generation, moderation and ranking,…
A data engineer built a local-first memory layer for AI agents using SQLite and FTS5 full-text search instead of vector embeddings, claiming it matches recall quality while eliminating cloud latency, opaque storage, and …
A developer tested whether YAMNet, Google's audio classification model, can detect unseen sudden sounds in real time without prior specification. The best configuration achieved 45.8% recall (22 of 48 events), but the re…
A developer shared in a DevOps discussion that an AI code reviewer on a Node and React project repeated style complaints, commented on outdated code, and missed a null check that caused a crash, requiring full human revi…
A developer built Sleep Guardian, an edge AI system on Raspberry Pi that uses Whisper-tiny and Temporal Convolutional Networks to classify sleep sounds like snoring and apnea locally without sending audio to the cloud. T…
LingBot-Map, a feed-forward 3D foundation model for streaming scene reconstruction, achieves state-of-the-art performance at approximately 20 FPS on 518×378 resolution over sequences exceeding 10,000 frames. The model's …
Researchers introduce Visual Concept Inference from Sets (VICIS), a task that evaluates vision-language models' ability to infer shared concepts from example images and apply them to new inputs. State-of-the-art VLMs per…
OpenAI Whisper is the most capable open-source speech-to-text engine available in 2026, supporting 99+ languages with near-human accuracy, according to OpenAI. The model, trained on 680,000 hours of multilingual data, ru…
A new book, 'Time Series with PyTorch' by Graeme Davidson and Lei Ma, offers a practical guide for time-series forecasting using PyTorch, covering evaluation, architectures, and tasks beyond point forecasting. The book's…
Apple researchers have developed an efficient unlearning framework that reduces computational costs by up to ~50% by identifying and excluding low-influence data points before the unlearning process, according to a paper…
On-device text-to-speech engines are approaching cloud quality in spring–summer 2026, led by Supertonic's 99M-parameter model running 167× real-time across 31 languages, while cloud APIs consolidate the STT→LLM→TTS pipel…
Roboflow released a multi-model auto-labeling workflow for segmentation that combines SAM 3, Google Gemini, and OpenAI's GPT to produce pixel-perfect masks via a 2-of-3 consensus rule, eliminating uncorrelated errors and…