Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Agentic AI and generative AI differ in that agentic systems perceive their environment, plan multi-step action sequences, and execute those plans with minimal human intervention, while generative AI produces new text, im…
Researchers submitted SpeakerMem-R1, a speaker-centered dual-track memory system for multi-party dialogue, to arXiv on 22 Sep 2026. SpeakerMem-R1 achieved 62.33% on the publicly reported EverMemBench leaderboard from Eve…
A developer argues that the common Airflow-versus-Dagster orchestration debate misses the real failure mode in ML pipelines: task-centric DAGs treat models and datasets as opaque task outputs rather than versioned, linea…
A March 2, 2026 arXiv paper by Aditya Anirudh Jonnalagadda proposes the b-posit, a bounded posit format that caps the regime field at 6 bits and reports 79 percent lower power consumption, 71 percent smaller area, and 60…
Geely Automobile Holdings unveiled a charging system in Ningbo, China, on Wednesday that it claims can replenish an electric vehicle from 10% to 70% in 4.5 minutes and from 10% to 97% in 8 minutes and 40 seconds, using a…
Tether's AI research arm QVAC released Genesis III, a 191.43 billion-token STEM training dataset spanning 19 domains, freely available on Hugging Face under a CC-BY-NC 4.0 license. Models trained exclusively on Genesis I…
A study of "jev-as-a-judge" compares a decision-only LLM judge against sixteen generative approaches, testing whether a cheaper first-pass judge can flag when stronger evaluation is needed. The work targets inference cos…
Researchers presented memristive SVD (MSVD), a singular value decomposition system built on compute-in-memory memristor chips and enabled by a selective representation enhanced architecture (SREA), reporting order-of-mag…
An IETF internet draft on privacy-preserving federated learning in multi-tenant agent systems argues that enterprise AI contracts quoting a differential privacy budget as a single epsilon and delta figure are not, on the…
An AI system called Astra produced an 8-page proof of the Liouville version of the Goldbach conjecture, which states that every positive even number greater than 2 is the sum of two positive numbers with Liouville value …
Engineer Richard Emate argues that a software library functions as a basis, application code as its coefficients, and refactoring as a change of basis, using compression and sparse coding to define what qualifies as a ge…
Gartner predicts organizations will abandon 60% of AI projects by the end of the year due to a lack of data readiness, and almost two-thirds of organizations are unsure whether they have the right data management practic…
NVIDIA released the Nemotron-H family of hybrid Mamba-Transformer models, including 8B models (Nemotron-H-8B-Base, Nemotron-H-8B-Instruct, Nemotron-H-8B-VLM) and 47-56B models (Nemotron-H-47B-Base, Nemotron-H-56B-Base, N…
A computer engineering graduate with a traditional embedded systems background is asking whether knowledge of C and C++, Linux networking, memory management, multithreading, synchronization, and interrupts is sufficient …
Chris, Lead Architect of the ANC-Loss-Tensor project, released an official PyTorch implementation of the Asymptotic Non-Closure (ANC) loss function, L_ANC, with the paper registered under DOI 10.17605/OSF.IO/Q8G9S. The r…
Sequen, a ranking startup founded in 2024 by Zoe Weil and Mo Afshar, raised a $90 million Series B in September after developing a technology it calls "Recursive Ranking Intelligence," the company told The Stack. Sequen …
Uber detailed a feature logging framework it built to keep the features used at inference identical to those used in training, addressing online/offline mismatches such as language values formatted as "en" at serving ver…
Cosine similarity measures how alike two vectors are by dividing their dot product by the length of both vectors, cancelling out magnitude and leaving only the angle between them, according to the third installment of a …
Jayanth, a freelance full-stack and AI automation developer and B.Tech student (Class of '26), introduced himself to the DEV Community and described building a CNN-BiLSTM deep learning gesture recognition system for the …
A new paper, "StableVQ: Practical Guidelines for Stable Vector-Quantized Tokenizer Training," addresses training stability in vector-quantized visual tokenizers, which underpin autoregressive and masked image generation …