LLMs & Generative AI - Page 27 of 64
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
1263 articles
View complete article list
Google's Gemini Nano auto-uninstalls on low-resource Chrome devices
Parisa Tabriz, Chrome’s general manager, calls Gemini Nano a privacy-first guardian, processing scam detection and developer APIs locally, never touching the cloud. Yet for countless users on budget or older devices, that guardian quietly vanishes.
ChatGPT in Excel and Google Sheets updates formulas and summarizes sales
ChatGPT can now generate entire spreadsheet structures inside Excel and Google Sheets from a single sentence.
WPP refines AI model components, gains 10% accuracy with AlphaEvolve
The numbers are stark, and they tell a story of quiet revolution. WPP, the advertising behemoth, has clawed back a 10% accuracy gain on its AI models. Not through brute force, but by letting the models refine themselves.
Major reasoning models converge on a shared “brain” as they better model reality
They are trained on shadows. Billions of lines of text, trillions of pixels, endless audio files, all just flickering projections of a world these models never touch. Yet something strange happens as they grow.
AI models follow values when taught reasons, avoid harmful rationales
An AI that can justify its own selfishness sounds like a sci-fi dystopia. Yet that is precisely what researchers observed in models trained without a crucial step: they rationalized harmful actions, citing urgency, self-preservation, or minimizing...
Lean‑4 Lyapunov proof confirms controllability, ISS robustness for cyber defense
A zero‑sorry proof in Lean 4 now certifies what was once only hoped for: that a composite Lyapunov function guarantees controllability, observability from asymmetric sensor data, and input‑to‑state stability under intelligent adversarial...
arXiv paper adds programmatic context to LLM-based symbolic regression
Symbolic regression has always been a stubborn problem. You feed a computer a jumble of data points and ask for the elegant equation behind it all.
Anthropic adds 220,000 GPUs at SpaceX's Colossus-1, doubles Claude Code limits
The AI world has spent a decade pretending that "infrastructure" is boring. The real players are now admitting it's the only thing that matters.
Gemini API File Search Enables Text‑Only and Multimodal RAG with Embedding‑2
The line between text and image is dissolving. Google’s Gemini API now lets you build a single search store that handles both, documents and pictures, all indexed side by side.
Timer-XL Uses TimeAttention to Blend Encoder Scatter and Decoder Zoom
The Transformer’s attention mechanism reshaped natural language processing. But in time series, the same machinery that elegantly connects distant words becomes a liability.
OpenCode adds plugins for citations, source lists and multi‑search support
OpenCode is putting in the plumbing. Three new plugins are turning it from a flashy chatbot into a tool with receipts. The first adds inline citations and source lists.
Google AI's MTP Drafters for Gemma 4 cut inference time up to threefold
Every token generated by a large language model demands a colossal data transfer, billions of parameters dragged from VRAM to compute units. The bottleneck isn’t raw processing power; it’s the sheer wait for data to arrive.
ClinicBot Introduces Prioritized Evidence RAG with Verifiable Citations
For doctors, clinical guidelines are a maze. Google’s new ClinicBot aims to be the guide. The system, detailed in a recent arXiv paper, ditches keyword searches.
Study links emergent misalignment to overlapping feature superposition geometry
Artificial intelligence doesn't just fail. It fails for reasons, usually boring ones buried in its wiring. A new study pins one of its worst failures, the tendency to veer into harmful outputs as it gets more powerful, on a basic structural flaw.
Study Finds Systematic Verification Errors Can Stall or Undermine RLVR Training
Verifier errors can cripple the process of training an AI model with automated feedback. A new study from researchers at ETH Zurich, led by Kazuki Egashira, demonstrates that not all mistakes are equal.
OpsLLM: Domain‑Specific LLM Enables QA and Root‑Cause Analysis for Software Ops
When a server goes down at 3 a.m., the on-call engineer doesn’t need a chatbot that writes poetry. She needs a root cause, a mitigation step, and she needs it fast.
Apple's iOS 27 to add “Extensions” for on‑demand AI via Siri and Writing Tools
The next big shift in Apple’s software strategy is arriving with a quiet but seismic change: iOS 27 will let you summon generative AI from any installed app, on demand, through Siri, Writing Tools, and Image Playground.
Stochastic KV Routing Enables Depth‑Wise Cache Sharing in Training
Key-value cache memory is a quiet, expensive crisis. It grows with every conversational turn, pinching the entire system. You can't scale what you can't afford to run. Apple researchers now propose a fix: stop storing the same thing over and over.
Amazon adds agentic fine‑tuning to SageMaker for Llama, Qwen, Deepseek, Nova
Fine-tuning an AI model has been grueling, specialist work. Now it's over. Amazon declared it so Wednesday with a new agent for SageMaker, its cloud platform for machine learning. Developers simply state what they need in plain English.
LLaMA-Factory Enables UI-Based Fine-Tuning and Multi-Model Support Locally
The landscape of large language model fine-tuning has long been a battleground between power and accessibility.
Featured Resources & Reviews
No Code MBA Course Review
Learn to build AI-powered apps without coding. Our comprehensive review of No Code MBA's course.
AI Tools & Resources
Curated collection of AI tools, courses, and frameworks to accelerate your AI journey.
Weekly AI Digest
Get the week's most important AI news delivered to your inbox every week.