Research & Papers
The KV Cache Compression Race: TurboQuant vs OSCAR vs EpiCache
The article compares three KV cache compression techniques—TurboQuant, OSCAR, and EpiCache—that address the memory...
MarkTechPost
Research & Papers · Other Companies
Research & Papers
In game theory, generalists sometimes win out over specialists
Researchers have discovered that a previously overlooked class of algorithms performs significantly better than...
MIT News AI
Research & Papers
Collecting robot training data is dirty, unglamorous work. Some AI labs are already paying XDOF to do it.
The article discusses the challenge of collecting training data for physical AI systems, noting that while large...
TechCrunch AI
Research & Papers
Could AI tell you where you left your keys?
Researchers have developed a new spatial memory system for robots that efficiently captures and recalls details about...
MIT News AI
Research & Papers
How to Build Memory-Efficient Transformers with xFormers Using Packed Sequences, GQA, ALiBi, SwiGLU, and Causal Attention
This article demonstrates how to implement xFormers, a toolkit for building memory-efficient Transformer models on GPUs...
MarkTechPost
Research & Papers
Meet Qwen-RobotSuite: Three Embodied AI Models for VLA Manipulation, Video World Modeling, and Navigation
Alibaba's Qwen team introduces Qwen-RobotSuite, a collection of three embodied AI models for robotic applications...
MarkTechPost
Research & Papers
How to Build a Parsing Pipeline with Docling Parse for Layout-Aware Document Intelligence
This tutorial demonstrates how to use Docling Parse to analyze PDF documents with layout-aware document intelligence,...
MarkTechPost
Research & Papers
Meet Flash-KMeans: An IO-Aware, Exact K-Means That Runs Over 200× Faster Than FAISS on GPUs
Flash-KMeans is an open-source, IO-aware GPU implementation of k-means clustering that achieves significant performance...
MarkTechPost
Research & Papers
A Coding Hands-On on FineWeb for Streaming, Filtering, Deduplication, Tokenization, and Large-Scale Web Corpus Analytics
This tutorial provides a hands-on guide to working with the FineWeb dataset, demonstrating how to stream and analyze...
MarkTechPost
Research & Papers
New AI model called "Count Anything" does exactly what it says, and that's harder than it sounds
"Count Anything" is a new AI model that can count objects in any type of image using text prompts, cutting error rates...
The Decoder
Research & Papers
A Coding Implementation on Spatial Graph Neural Networks for Urban Function Inference Using city2graph, OSMnx, and PyTorch Geometric
The article presents an end-to-end spatial graph learning pipeline using city2graph, OSMnx, and PyTorch Geometric to...
MarkTechPost
Research & Papers
When it comes to predicting people’s preferences, it pays to consider “the power of three”
MIT researchers have developed a significant advancement to random utility models, a theory nearly a century old that...
MIT News AI
Research & Papers
How memory tools can make AI models worse
Recent research reveals that memory systems integrated into AI models can actually reduce performance and promote...
TechCrunch AI
Research & Papers
ClawHub Security Signals: A Coding Guide to End-to-End Security Signal Analysis and Verdict Classification on the AI Skills Dataset
This tutorial explores the ClawHub Security Signals dataset to analyze how security scanners assess AI skills,...
MarkTechPost
Research & Papers
Researchers pinpoint why larger language models pick up skills that small ones miss
Researchers discovered that small language models fail at rare tasks because frequent tasks overwrite their learned...
The Decoder
Research & Papers
A Hands-On Coding Tutorial on Qualcomm AI Hub Models for Classification, Object Detection, and Hardware-Aware Deployment
This tutorial demonstrates how to use Qualcomm AI Hub Models to run inference tasks including MobileNet-V2...
MarkTechPost
Research & Papers
NVIDIA AI Releases Dynamo Snapshot: A CRIU-Based Fast Startup System for AI Inference on Kubernetes
NVIDIA has released Dynamo Snapshot, a system that uses CRIU and cuda-checkpoint tools to enable fast startup of vLLM...
MarkTechPost
Research & Papers
Building a Semantic Search Engine and Open-Status Classifier over the ResearchMath-14k Dataset
This tutorial demonstrates a complete NLP pipeline for research mathematics using the ResearchMath-14k dataset,...
MarkTechPost
Research & Papers
NSF renews support for MIT-led AI and physics institute, expanding a new model for discovery
The National Science Foundation has renewed funding for IAIFI (Institute for Artificial Intelligence and Fundamental...
MIT News AI