Machine learning news — deep learning, reinforcement learning, neural architecture search, diffusion models, and new ML frameworks and libraries.
Researchers have developed Memory-as-Action (MemAct), a framework that treats working memory management as learnable policy actions for large language models, enabling joint optimization of information retention and task…
A final-year computer science student from West Bengal, Priyanshu, built NiDaan, an offline AI diagnostic assistant for ASHA health workers in rural India. The system operates without internet, accepts symptoms in Hindi …
A three-woman research team at the Birla Institute of Technology Mesra in Ranchi has trained artificial intelligence models to detect and analyze lunar craters, with funding from the Indian Space Research Organisation. T…
An engineer built a four-tier memory architecture for AI agents, implementing working, semantic, episodic, and procedural memory systems from scratch. The system enables agents to retain context across sessions, recall p…
Large language models like ChatGPT, Gemini, and Claude operate as next-token prediction machines, generating text by repeatedly calculating probability distributions over their vocabulary for what token should follow a g…
Komi-learn, a new tool for coding agents, automatically learns a user's coding style, stack, and fixes from sessions and recalls them in future sessions without commands. The tool works with Claude Code and Codex, distil…
Sina Rajaeeian, a software engineer and machine learning master's student at KTH Royal Institute of Technology in Stockholm, has launched an interactive 3D portfolio website showcasing his work across AI/ML systems, full…
Open-weight AI models now trail proprietary state-of-the-art systems by an average of four months in performance, according to Epoch's internal capability metric. The Epoch Capability Index, which aggregates benchmark re…
Researchers have developed a conditional scaling law that predicts optimal large language model architectures for inference efficiency, training over 200 models from 80 million to 3 billion parameters. The study found th…
Researchers found that using diverse ensemble monitoring—combining multiple small monitors built from a single base model—outperforms both repeated sampling of the same monitor and a single, more costly monitor for detec…
A developer benchmarked four lightweight transformer models—DistilBERT, MobileBERT, TinyBERT-6L, and TinyBERT-4L—against traditional ML baselines for fault detection, finding that TinyBERT-4L achieved 87.8% F1 with 55 MB…
Hermes Repo Dojo, a new local AI developer tool, transforms any public GitHub repository into a guided onboarding academy by enabling the Hermes Agent to learn, create reusable skills, and safely contribute code. Unlike …
Trajectory, in collaboration with UC Berkeley Sky Lab and Anyscale, released a concurrent multi-LoRA training stack for continual learning that achieves a 2.81× end-to-end experiment-throughput gain over single-tenant re…
The W3C draft specification WebNN defines a graph-based neural network API that enables JavaScript to submit static model graphs to native OS runtimes, according to a November 21, 2025 blog post on ziade.org. Browsers ma…
Google Chrome is deploying on-device machine learning warnings on Android to flag deceptive or spammy notifications, displaying the sending site, a warning message, and options to unsubscribe or allow future alerts. The …
A developer released SnapName, a Mac app that uses a local AI model to automatically rename screenshots with descriptive filenames. The app watches a designated folder and generates three AI name suggestions for each new…
A developer has built a personal AI newspaper using the Hermes agent, a self-improving AI system that features a built-in learning loop called GEPA (Generate, Evaluate, Prune, Accumulate). The agent retains memory from p…
A 10-year AI engineer transitioning from FAANG to finance seeks book recommendations and advice for building a B2C product in net worth tracking, agentic finance, and LLM-driven investing. The engineer, who has built alg…
A new open-source repository, "AI for Beginners Visual," offers simple diagrams and everyday analogies to explain core artificial intelligence concepts like tokens, embeddings, and neural networks to non-technical reader…
The open-text-to-speech (TTS) landscape has shifted dramatically between 2024 and mid-2026, with open-weight models now rivaling proprietary APIs from ElevenLabs, Google, and OpenAI in quality while offering lower costs …