Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
A 2026 study by Daeyoung Roh and Donghee Han of 12,000 paired agent trajectories across HotpotQA, 2WikiMultiHopQA, and MuSiQue found that agentic RAG failures often stem from procedural errors: on MuSiQue, 617 of 1,000 t…
AMD's RDNA 4m integrated graphics have been merged into the Mesa 26.3 open-source driver stack under the GFX1171 GPU target, marking its official support in the open-source ecosystem. The RDNA 4m, which sits between RDNA…
Google Cloud announced the general availability of Autonomous Embedding Generation and AI.SEARCH in BigQuery, along with the public preview of Hybrid Search, to simplify unstructured data analysis. AI.SEARCH now delivers…
Anthropic is recruiting a custom-silicon team for Claude, with a Silicon Engineer listing updated August 4, 2026 and a chip-design reinforcement-learning role updated August 3, offering US base salaries of $320,000–$485,…
Roboflow published a tutorial for building an AI video car damage inspector that uses an RF-DETR model, ByteTrack tracking, and Gemini to generate digitally signed PDF reports. The system timestamps video fingerprints be…
A Stanford- and Arc Institute-led team reported in Science on August 6 that genome language models generated complete bacteriophage genomes, 16 of which proved viable in laboratory tests. The researchers synthesized and …
Perforated AI released a perforated version of ResNet-18 on Hugging Face that matches ResNet-34's performance while maintaining ResNet-18's smaller size and compute efficiency, using dendritic optimization technology. Th…
A study by French economists, including Nobel laureate Philippe Aghion, found that a predictive machine-learning model for aeration control in French wastewater treatment plants reduced electricity consumption by 5.4%, c…
Teams running Spot-heavy Kubernetes configurations save 77% on compute on average, compared to 59% for mixed fleets, according to the 2025 Cast AI Kubernetes Cost Benchmark. AWS gives a 2-minute interruption notice, whil…
JetBrains Academy's July digest announces new learning resources, including a Gen AI Engineer Bootcamp by Ardit Sulce for building AI applications and agents, a pandas course with portfolio projects, and free access to K…
In a benchmark of 27 trials with a frozen coding task, the NVIDIA GB10 Grace Blackwell appliance in the ASUS Ascent GX10 served an 80B coding model 1.71 times faster per token than the AMD Ryzen AI MAX+ PRO 395 in the HP…
Vector quantization and product quantization offer an alternative to scaling-based neural network quantization, achieving high compression ratios by mapping high-dimensional vectors to a finite codebook. For example, wit…
Microsoft has open sourced code-testing-generator, a polyglot unit-test agent in the MIT-licensed dotnet/skills repository that reads a repository to detect language, test framework, conventions, and build/test commands …
UC Berkeley, ICSI, and LBNL researchers introduced ARBITRAGE, a step-level speculative decoding framework that uses a lightweight router to dynamically choose between draft and target model steps, reducing inference late…
Researchers at Apple trained a 1.7B-parameter categorical flow model on 2.1T tokens and self-distilled it into a Categorical Flow Map (CFM) that generates text in as few as 4 inference steps while maintaining near-data-l…
NOAA is moving its weather forecasting operations from traditional supercomputers to Google Cloud, a transition expected to cut forecast generation time from hours to seconds and be completed by December 2027. Google Dee…
Alibaba released Qwen3.8-Max, a 2.4-trillion-parameter Mixture-of-Experts model that activates about 95 billion parameters per query, available now via the Alibaba Cloud API and in Roboflow Playground, with open weights …
Roboflow published a guide on building an AI defect detection system using computer vision, demonstrating the process with ceramic tile defect data from Roboflow Universe. The system can automatically detect defects in r…
RAGproof released an open-source tool, rag-staleness-check, that performs read-only staleness, orphan, duplicate, and retrievability checks on a single pgvector, Qdrant, or Chroma index. The tool, installable via pip for…
Amazon SageMaker Python SDK v3.17.0 now integrates generative AI inference recommendations, allowing users to benchmark endpoints, generate deployment recommendations, and deploy optimized configurations directly from no…