LLMs & Generative AI - Page 11 of 48
Latest breakthroughs in large language models and generative AI shaping the future of artificial intelligence and machine learning.
951 articles
View complete article list
GPT‑5.5 and GPT‑5.5‑Cyber Scale Trusted Access for Cyber Defense
Why does this matter now? For the first time, a single family of models is being positioned to serve every layer of the cyber‑defense ecosystem. While the tech is impressive, the rollout is deliberate.
Google's Gemini Nano auto-uninstalls on low-resource Chrome devices
If you open Chrome on a Windows or macOS desktop today, a 4 GB Gemini Nano model is probably already sitting on your hard drive. Google slipped the file into the browser in 2024, and most users never see a prompt that it’s there.
ChatGPT in Excel and Google Sheets updates formulas and summarizes sales
AI isn’t always about flash‑forward features; most people just want tools that smooth out the daily grind. While the hype around “next‑gen” models rolls on, OpenAI has slipped something practical into its latest release.
WPP refines AI model components, gains 10% accuracy with AlphaEvolve
AlphaEvolve has moved out of the lab and into the backbone of Google’s AI infrastructure. While the system started as a Gemini‑powered coding agent, it now tackles hardware‑level design, software compaction and commercial workloads.
Major reasoning models converge on a shared “brain” as they better model reality
Why does it matter that wildly different AI systems might be thinking alike? While the tech is impressive, the question cuts to the heart of what we call “understanding.” In 2024, MIT researchers unveiled solid evidence that every major reasoning...
AI models follow values when taught reasons, avoid harmful rationales
AI labs are wrestling with a stubborn problem: models often follow written “Model Specs” but stumble when they encounter scenarios those specs never covered.
Lean‑4 Lyapunov proof confirms controllability, ISS robustness for cyber defense
Why does this matter for security operations centers? Because the paper arXiv:2605.03034v1 proposes a new way to harness large‑language‑model agents for autonomous cyber defense.
arXiv paper adds programmatic context to LLM-based symbolic regression
The arXiv preprint 2605.03101v1, titled *Programmatic Context Augmentation for LLM‑based Symbolic Regression*, tackles a problem that sits at the heart of many scientific workflows: turning raw measurements into concise mathematical formulas.
Anthropic adds 220,000 GPUs at SpaceX's Colossus-1, doubles Claude Code limits
Anthropic is moving the bulk of its compute to SpaceX’s Colossus‑1 data center. The deal gives the AI firm access to more than 300 megawatts of power and over 220,000 NVIDIA GPUs, with the hardware slated to be operational within a month.
Gemini API File Search Enables Text‑Only and Multimodal RAG with Embedding‑2
Why does this matter? Because building a retrieval‑augmented generation (RAG) pipeline used to require stitching together chunking, embedding and indexing yourself. Google’s File Search tool for the Gemini API now does that work for you.
Timer-XL Uses TimeAttention to Blend Encoder Scatter and Decoder Zoom
Why does this matter? Because forecasting models that only work on fixed windows are hitting a ceiling.
OpenCode adds plugins for citations, source lists and multi‑search support
Why does this matter? Because OpenCode’s core AI coding agent can now be extended with add‑ons that turn a basic assistant into a more versatile development partner.
Google AI's MTP Drafters for Gemma 4 cut inference time up to threefold
Here’s the thing: Google’s latest Gemma 4 update promises up to three‑times faster inference, and the company insists the quality stays intact.
ClinicBot Introduces Prioritized Evidence RAG with Verifiable Citations
Why does this matter? Because clinicians can’t afford guesswork when a patient’s life hangs in the balance.
Study links emergent misalignment to overlapping feature superposition geometry
Emergent misalignment has become a focal point for AI safety researchers. Why does fine‑tuning a language model on a narrow, seemingly harmless task sometimes unleash harmful behavior?
Study Finds Systematic Verification Errors Can Stall or Undermine RLVR Training
Reinforcement Learning with Verifiable Rewards (RLVR) promises to sharpen large language models’ reasoning by tying training signals to ground‑truth answers.
OpsLLM: Domain‑Specific LLM Enables QA and Root‑Cause Analysis for Software Ops
Microsoft Research is putting large language models to work on a problem that has long plagued hyperscale operators: turning noisy incident reports into clear, actionable guidance.
Apple's iOS 27 to add “Extensions” for on‑demand AI via Siri and Writing Tools
Apple is gearing up to roll out iOS 27 with a new “Extensions” layer that lets third‑party apps plug into the system’s AI functions when users invoke Siri, the built‑in writing assistant, or the visual sandbox tool.
Stochastic KV Routing Enables Depth‑Wise Cache Sharing in Training
Stochastic KV routing has been generating buzz among researchers working on large language models, promising a more flexible way to manage the key‑value caches that sit at the heart of transformer inference.
Guide to Getting Claude Code to Validate Its Own Output Automatically
Why does it matter if an LLM can check its own output? While the idea of “making Claude verify its own work” circulates in developer forums, the practical steps remain fuzzy.
📚 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.