LLMs & Generative AI - Page 8 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
Basis Spline Decoupling Enables Compression of Transformer Models
Why does compressing large transformers matter? Because the sheer size of modern Vision and Swin Transformer models makes deployment costly.
LLM Retrieves Median 2020 Inflation Expectation, Drowning Prompt Guidance
In a new paper titled *Can LLMs Mimic Household Surveys?*—co‑authored with Ami Dalloul of the University of Duisburg‑Essen—the authors turn a critical eye to what happens when large language models stand in for real respondents.
QuickReduce FP4 delivers ~4.1× speedup over RCCL at TP=4 for large messages
Large language models run on billions of parameters, and inference almost always leans on tensor parallelism to split the workload across several GPUs.
Study Uses SHARP and New Error Framework to Assess PHRs in Health AI
Personal health records promise patients a clearer view of their own data, but the sheer volume and medical jargon can make the files hard to navigate.
Google's Gemini 3.5 Flash, pricier, adds 11 Omniscience points hallucinations 61%
Google DeepMind rolled out Gemini 3.5 Flash this week, positioning it as the fastest model in its class—more than 280 output tokens per second, according to the company.
Alibaba launches Qwen3.5‑LiveTranslate‑Flash: 60‑language translation in 2.8 s
Simultaneous interpretation has always been a tough AI problem. You’re asking a model to translate speech before the speaker even finishes a sentence, and every extra millisecond chips away at the illusion of real‑time dialogue.
I/O 2026 unveils Gemini Omni for universal creation, Gemini 3.5 Flash debut
Google I/O 2026 put a new spin on the company’s AI roadmap. The headline? Gemini Omni, a model that “can create anything from any input,” starting with video, and Gemini 3.5 Flash, the first of a family that pairs “frontier intelligence with...
EKS Hosts Multistage Multimodal Recommender; DLRM Personalizes Rankings
Why does this matter? Because building a recommender that can juggle images, tabular data, and real‑time user signals is no longer a research exercise—it’s a production challenge.
Gemini 3.5 Flash Enhances Web UI, Graphics and AI Studio Animations
Google unveiled Gemini 3.5 today, positioning it as the next step in the company’s “frontier intelligence with action” roadmap.
Fresh Web Data Grounds LLMs, Highlighting RAG's Production Limits
Why does this matter? Companies are betting on large language models (LLMs) to answer everything from support tickets to sales queries, yet most LLMs stop learning at a fixed knowledge cutoff.
ANNEAL lets neuro‑symbolic agents patch knowledge graphs without weight changes
LLM‑driven agents can bounce back from a single slip‑up, yet they often stumble over the same mistake when the underlying procedural knowledge stays broken.
Claude Cowork: Guide to Turning Q1 Sales Data into a Structured Word Report
Most people treat Claude like a smarter search engine: you type a prompt, skim the answer, copy it somewhere, then start over. That loop works for quick look‑ups, but it stalls when a project spills across dozens of files, spreadsheets and slides.
Activation steering reveals latent bias in LLMs, reinjection restores decisions
Why does this matter? Because the fairness of AI decisions is often judged only by what the model says, not what it thinks.
95% of task‑specific generative AI pilots never reach production
The headline is stark: 95 % of task‑specific generative‑AI pilots never make it past the demo stage. Six months after a shiny proof‑of‑concept, many projects are simply abandoned.
5 Practical Uses of Local Language Models Highlight Code‑First Approach
The first time you type `ollama run llama3.2` and watch a 7‑billion‑parameter model spin up on your own laptop, something clicks. No API key pops up, no billing dashboard looms, and nothing leaves your computer.
SkillSmith extracts fine-grained boundaries so agents run only needed components
Why do most LLM‑based agents still lug around entire skill libraries? In many systems a skill is dropped into the reasoning loop as a monolithic prompt, even when only a tiny fragment is needed for the current task.
Quantized LLMs Show Emerging Bias, Masking Gradual Degradation
Why does shrinking a model matter beyond speed? While developers tout lower‑bit inference as a cheap fix for cloud and edge constraints, the fairness side effects have been largely invisible.
AgentStop cuts GPU power, heat and battery drain by ending AI agents early
Why does this matter? Autonomous agents built on large language models are moving beyond single‑prompt answers to tackle multi‑step jobs—coding snippets, navigating web pages, answering complex queries.
Vercel Labs launches Zero, a systems language for AI agents to read and ship
Zero is Vercel Labs’ first foray into a language built for machines, not humans. Released on May 15, 2026, the experimental systems language—currently at v0.1.1—ships native binaries under 10 KiB and uses the .0 file extension.
Enterprise‑grade AI platform merges chatbot, voice, video; customizable via APIs
Unite.AI’s editorial team sticks to strict standards, and we’ll be upfront: we may earn compensation when readers click our product links.
📚 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.