Researchers baked 3x inference speedups directly into LLM weights — without speculative decoding

Researchers from the University of Maryland, Lawrence Livermore, Columbia and TogetherAI have developed a training technique that triples LLM inference speed without auxiliary models or infrastructure ...

From ‘Optical Synapses’ to a ‘Photonic Brain’ — Integrated Photonic Neural Networks Toward Low-Power General-Purpose

SHANNON, CLARE, IRELAND, February 5, 2026 /EINPresswire.com/ -- A new publication from Opto-Electronic Technology; DOI ...

Micron's New SSDs Are Gamechangers For AI

SSDs represent a robust growth vector for Micron Technology as memory demands in AI data centers show no signs of stopping.
Live Science on MSN

MIT designs computing component that uses waste heat as a form of information

Proof of concept uses passive components to redirect heat across a chip, allowing temperature patterns to be used for data processing.
IEEE

Wireless Distributed Matrix-Vector Multiplication Using Over-the-Air Computation and Analog Coding

Abstract: In this paper, we propose an over-the-air (OTA)-based approach for distributed matrix-vector multiplications in the context of distributed machine learning (DML). Thanks to OTA computation, ...
Business Wire

SiFive’s New RISC-V IP Combines Scalar, Vector and Matrix Compute to Accelerate AI from the Far Edge IoT to the Data Center

SANTA CLARA, Calif.--(BUSINESS WIRE)--Further expanding SiFive’s lead in RISC-V AI IP, the company today launched its 2nd Generation Intelligence™ family, featuring five new RISC-V-based products ...
GitHub

GLSL matrix vector multiply operator (*) overloads do not work anymore

import glsl; [shader("fragment")] void fragment_main() { mat4 matrix = mat4(1.0); vec4 vector = vec4(1.0); vec4 result0 = matrix * vector; vec4 result1 = matrix ...
C&EN

Thermodynamics Analysis of a Reaction-Diffusion Matrix Multiplication Computing Unit under the Linear Non-Equilibrium Regime

Creative Commons (CC): This is a Creative Commons license. Attribution (BY): Credit must be given to the creator. Implementations of matrix multiplication via diffusion and reactions, thus eliminating ...
GitHub

multiply_matrix_vector.h

/// @brief Module for handling the matrix-vector multiplication as a part of solving the 1d PDE for heat diffusion. /// Options are: /// 1. 'manual' : using explicit triple loop for matrix-vector ...