New KV cache compaction technique cuts LLM memory 50x without accuracy loss

MIT researchers developed Attention Matching, a KV cache compaction technique that compresses LLM memory by 50x in seconds — ...

Black Forest Labs' new Self-Flow technique makes training multimodal AI models 2.8x more efficient

This efficiency makes it viable for enterprises to move beyond generic off-the-shelf solutions and develop specialized models ...

Fiber setup compresses mid-infrared pulses to 187 femtoseconds using just 80 watts

Ultrashort mid-infrared (mid-IR) laser pulses are essential for applications such as molecular spectroscopy, nonlinear microscopy, and biomedical imaging, but their generation often relies on complex ...

(CNW) MulticoreWare and Small Pixels Announce AI-Powered Integrated Video Optimization Solution to Dramatically Reduce Bandwidth and Infrastructure Costs

MulticoreWare, Inc., and Small Pixels today announced a strategic technology collaboration merging two powerful optimization solutions: MulticoreWare's Ultraziq and Small Pixels' SPAIQ Stream. The ...
Tom's Hardware on MSN

Microsoft debuts DirectStorage 1.4 at GDC 2026, with Zstandard compression and GACL

DirectStorage 1.4 brings along key upgrades to the API, including support for Zstandard compression as well as CreatorID for improved GPU scheduling.
The New York Times

Artificial Intelligence

Valued at $1.6 billion, a tiny start-up called Axiom is building A.I. systems that can check for mistakes. By Cade Metz Following rivals like Amazon and OpenAI, Microsoft is upgrading its artificially ...
WinBuzzer

Gemini Embedding 2 Unifies Text, Images, Video in One Model

Google has launched Gemini Embedding 2, its first natively multimodal embedding model supporting text, images, video, audio, ...