Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers primarily from Tsinghua University's THU-NICS group introduced Cache-to-Cache (C2C), a framework that lets large language models exchange key-value caches directly instead of text, delivering accuracy gains o…
LinearSolveBench launched as a benchmark measuring how well AI models and harnesses write fast, accurate, general numerical solvers for large sparse linear systems in C. On the FLASH magnetic-diffusion leaderboard, Fable…
Google Fellow John Platt outlined a method for accelerating empirical discovery by mapping scientific challenges into scoreable tasks and applying LLM-guided tree searches, work he says spans applications from fusion tim…
Block posted an on-site Staff Machine Learning Engineer (Modeling), Support role in Seattle, WA, requiring 10+ years of machine learning, applied AI, or product ML experience and skills in RAG, PyTorch, TensorFlow, deep …
A new research effort targets full-pipeline FP8 reinforcement learning for large language models, addressing the difficulty of maintaining stability across an FP8 RL pipeline even though FP8 quantization can accelerate R…
A Hacker News user posting as kooldeep7 asked the site's community when fine-tuning a small LLM is worth it, requesting accounts of the task, the model trained, and the results. The Ask HN thread, submitted 13 minutes be…
Unblocked reported that its calibrated judgment model Jev outperformed a cross-encoder reranker on agent memory selection, achieving higher precision and recall across 12,927 labeled question-note pairs drawn from 292 pr…
A developer released Shrewd, a project that distills labels generated by frontier LLMs into small, single-purpose classifiers that run efficiently on-device, and published results on public datasets used for such compari…
A practical guide demonstrates transfer learning with MobileNetV2 in Keras, using a frozen pretrained base plus a new Dense(5) output layer to classify 3,670 flower photos across five categories (roses, daisy, dandelion,…
A developer improved retrieval overlap on a legal-contract benchmark from 37.3% to 50.4% by retrieving at the sentence level and then returning the full paragraph each hit came from, using stela chunking, all-MiniLM-L6-v…
Sony Group Corporation filed two US patent applications, published 17 September 2026, describing an AI system that could remotely diagnose faults in a PlayStation or other connected device and use a large language model …
Laya released laya-coreml, an open-weight Core ML model that runs typed decision inference locally on Apple Silicon's Neural Engine, sustaining 49.1–50.0 decisions/s across three uncapped 600-step Snake episodes with zer…
A developer fine-tuned an 8B-parameter model on fewer than 250 security-classification examples and documented a set of known degradation patterns appearing together at small scale on consumer hardware. The fine-tuned mo…
Cohere launched North Small Translate, an open machine translation model with 218B parameters (25B active) and a 16k context length that the company says outperforms other open MT models as well as Google Translate and D…
Researchers at the USC Viterbi School of Engineering and Stevens Institute of Technology developed an automated AI system that analyzes multilingual online job postings to detect deceptive recruitment tactics linked to h…
Approximate nearest-neighbor vector search has collapsed into existing databases — Postgres via pgvector 0.5.0 (August 2023), Elasticsearch, OpenSearch, ClickHouse 25.8, MongoDB Atlas (GA December 2023) and Redis — rathe…
A 2026 survey of 518 global enterprise technology leaders conducted by Delphix found that synthetic data is widely viewed as a critical capability for maintaining data privacy while accelerating software development, yet…
STMicroelectronics is demonstrating four Physical AI applications at Embedded World North America, including a stable robotic hand and a virtual humanoid walking simulator built on NVIDIA Isaac Sim and Isaac Lab, using S…
The journal Nature Reviews Physics published a forward-looking review article by an international team of scientists on self-organizing physical networks that compute directly within hardware rather than running neural n…
The arrowspace graph-building system scaled to 100,000-dimensional data, processing 800 samples from the Dorothea biomarker dataset in 41-110 seconds with over 600× memory compression (610 MB to 1 MB), according to a per…