GitHub

RotorQuant: KV Cache Compression for LLMs

The butterfly bypass from the RotorQuant paper: TurboQuant applies a d×d Walsh-Hadamard Transform (butterfly network with log₂(d) stages across all 128 dimensions). PlanarQuant/IsoQuant apply ...

Google TurboQuant: Separating hype from reality

When Google unveiled TurboQuant on March 24, headlines declared the algorithm could slash AI memory use sixfold with zero ...
marktechpost

Researchers from MIT, NVIDIA, and Zhejiang University Propose TriAttention: A KV Cache Compression Method That Matches Full Attention at 2.5× Higher Throughput

Long-chain reasoning is one of the most compute-intensive tasks in modern large language models. When a model like DeepSeek-R1 or Qwen3 works through a complex math problem, it can generate tens of ...
The Del Norte Triplicate

Domain Cache Does This Next Mission Reunion

Domain Cache Does This Next Mission Reunion. Full mass line. Can niacin make you dashing to get stone? Persuaded him not sack all around! Worst poet ever? Enterprise distribution ...
marktechpost

An End-to-End Coding Guide to NVIDIA KVPress for Long-Context LLM Inference, KV Cache Compression, and Memory-Efficient Generation

In this tutorial, we take a detailed, practical approach to exploring NVIDIA’s KVPress and understanding how it can make long-context language model inference more efficient. We begin by setting up ...
Business Wire

Axip Receives Court Approval for Sale of Substantially All Assets to Service Compression, LLC

HOUSTON & FORT WORTH, Texas--(BUSINESS WIRE)--Axip Energy Services, LP and certain of its affiliates (collectively “Axip” or the “Company”) and Service Compression, LLC (“Service Compression”) today ...
PC Magazine

Nvidia, Intel Texture Compression Techs Cut VRAM Use Dramatically

Intel and Nvidia showed off their respective AI-powered texture-compression technologies over the weekend, demonstrating impressive reductions in VRAM use while maintaining texture quality, or even ...
TechSpot

Nvidia shows neural compression can cut VRAM usage from 6.5GB to 970MB

Forward-looking: Nvidia's latest push into neural rendering is not just unfolding on keynote stages, but also in follow-up technical briefings. A recent video released days after the DLSS 5 ...
IEEE

SLAP-CC: Set-Level Adaptive Prefetching for Compressed Caches

Abstract: Data prefetching and cache compression are well-studied techniques to reduce the impact of memory latency. Data prefetching predicts future memory accesses and prefills the cache with the ...
Hosted on MSN

Claude triumphs as Alibaba launches new AI model

Claude has emerged victorious in the AI Madness 2026 competition, defeating ChatGPT in a rigorous seven-round final that tested coding, creative writing, and complex reasoning. On the same day, ...
Hosted on MSN

Google’s new AI compression could cut demand for NAND, pressuring Micron

A new compression technique from Google Research threatens to shrink the memory footprint of large AI models so dramatically that it could weaken demand for NAND flash storage, one of Micron ...
IEEE

Enhancing Large Multimodal Models with Adaptive Sparsity and KV Cache Compression

Abstract: Large multimodal models (LMMs) have advanced significantly by integrating visual encoders with extensive language models, enabling robust reasoning capabilities. However, compressing LMMs ...