Le projet THINK

Researchers from the University of Texas at Austin and Intel Labs investigated chain-of-thought reasoning structures in large language models to identify and calibrate flawed reasoning pathways

Large Language Models are revolutionizing AI applications; however, slow inference speeds continue to be a significant challenge. Intel researchers, along with industry and university partners, are actively working to address this issue and accelerate the efficiency of LLMs. In a series of blog posts, Intel Researchers introduce several novel works, including a method that accelerates text generation by up to 2.7 times, a method that extends assisted generation to work with a small language model from any model family, and a technique that enables any small “draft” model to accelerate any LLM, regardless of vocabulary differences

Ramtin Davanlou, CTO of the Accenture and Intel Partnership, explores what it really takes for enterprises to move from AI experimentation to scaled success, revealing how data quality, strategic partnerships, and AI factories are reshaping the way businesses operate.

In this blog, we demonstrate how to build and run a PDF Summarizer Agent using Intel® XPU-optimized tooling for efficient inference, leveraging CrewAI’s orchestration, PyPDF2 for text extraction, and a T5-based model for summarization.

In an era dominated by the narrative that “AI equals GPUs,” a quiet revolution is underway—one that shows not all AI needs to be powered by expensive, scarce, and energy-intensive accelerators.

Adversarial examples can force computer-use artificial intelligence (AI) agents to execute arbitrary code. To aid AI researchers in evaluating robustness of agentic models, Intel Labs researchers open sourced an adversarial image injection proof of concept (PoC) against computer-use AI agents such as UI-TARS

Discover how Intel® Gaudi® accelerators and the llm-d stack improve large language model inference by decoupling Prefill and Decode stages. Learn how this approach reduces latency, enables smarter scheduling, and supports hybrid deployments across Intel Gaudi accelerators and NVIDIA GPU’s. Scalable, efficient, and flexible—this is next-gen LLM inference in action!

Intel Labs researchers have developed a new set of safety concepts for mobile and stationary robots that enhance the robot’s operating capabilities while ensuring the robot always maintains a safe state.

Intel and Fluid Inference optimized transformer models to run locally on Intel AI PCs, enabling private, real-time AI. Read on to discover how this breakthrough was achieved.

Intel PyTorch & SGLang Team proposed a high-performance CPU-only solution using the Intel® Xeon® 6 Processor at a fraction of the cost.