Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
IIT Bombay has launched an e-Postgraduate Diploma in Computer Science and Artificial Intelligence, targeting working professionals, with Great Learning as the EdTech delivery partner. The program blends online sessions w…
Databricks Inc. is positioning its platform as a 'control plane for context' to help retailers scale AI across stores, headquarters, and digital channels while maintaining governance over sensitive customer data, model c…
A new open benchmark from the Steel-predictor project shows that Anthropic's Claude Sonnet 5 leads large language models in predicting knife-steel properties from chemical composition, achieving a mean Spearman rank corr…
The KV cache has no standard ABI, with vLLM's FlashAttention backend alone reporting its cache shape as a four-dimensional tensor that varies by backend, attention variant, and model family, complicating disaggregated in…
Shoehorn, a new open-source tool from developer notactuallytreyanastasio, quantizes any language model to fit a user's available memory, using up to 99.99% of the budget. It solves a per-tensor mixed-precision assignment…
Confluent announced new AI capabilities in Confluent Cloud, including the general availability of Real-Time Context Engine, expanded time-series model support with IBM Granite and Google TimesFM models, updates to its fu…
A developer's claim that their MadriMed-VL-2B model beats Google's MedGemma 4B on the MedXpertQA-MM benchmark is methodologically unsound because the evaluation likely ran text-only without loading images, yielding a 21.…
Developer tsuyu122 released USAF, an open-source sparse fine-tuning method for Mixture-of-Experts (MoE) models, enabling fine-tuning on consumer GPUs with as little as 12 GB VRAM. The method, demonstrated on an AMD RX 67…
A cloud architect detailed a production playbook for multimodal AI pipelines, emphasizing reliability, p99 latency, and cost predictability over benchmark performance. The engineer shared a circuit-breaker pattern that a…
A CS and Mathematics student from Ujjain, India, built and deployed LARA, a lightweight computer vision API that runs YOLOv8 nano on a CPU-only server, eliminating the need for GPUs. The API, built with FastAPI, Supabase…
Arb'a Baloot, a free, ad-free digital version of Saudi Arabia's national card game Baloot, has been released with an AI opponent trained using reinforcement learning over millions of self-play rounds. The AI, built by a …
Rainbow, an AI-powered weather startup, has developed a system that predicts rain from satellite data every 10 minutes and outperformed AccuWeather in a flood test in Indonesia, according to a Forbes report by Sandy Cart…
A diagnostic experiment by TraceML maintainer found that Qwen3-1.7B LoRA training with TRL on a single T4 ran 17% faster with per-device batch size 4 and no gradient accumulation (238.2s) than with per-device batch size …
A 500-page M&A purchase agreement parsed by an autonomous document review agent in 94 seconds flagged zero high-risk anomalies, yet outside counsel later discovered an unhedged $14 million environmental liability carve-o…
Synthefy Inc. has raised $6.5 million in seed funding to expand its Structured Data Foundation Models (SDFMs), which are trained on numerical data rather than text. The round was led by Wing Venture Capital with particip…
Shri Ram College of Commerce (SRCC) and EY India launched AI-VIBE, a 12-month certification programme for undergraduate students from Semester 3 onwards, blending machine learning, Generative AI, AI agents, and responsib…
A practical guide outlines a learning path for AI in 2026, emphasizing an engineering-level approach over deep research. The author recommends starting with essential mathematics and PyTorch projects, then progressing to…
Lakshman Pandey has shipped a production retrieval-augmented generation (RAG) system for UK arts and culture clients that implements NIST AI Risk Management Framework controls. The system uses Claude and Voyage APIs with…
A former head of data and machine learning engineering argues that not every problem needs an AI agent, citing a case where replacing ML models with AI agents at a former company led to unreliable results, higher latency…
The ARM64 architecture will introduce the BBML3 feature with Linux 7.3, replacing the existing BBML2-noabort code for more efficient page table transitions and improved memory management. The kernel also includes workaro…