'Like gene editing for neural networks': MIT researchers used GPT-6 Astra to edit Qwen's open weights directly, boosting math performance

In the strongest example of "weight space programming", as the technique is called, Astra’s intervention raised Qwen3.5-0.8B’s MATH-500 score from 7.9% to 48.1% while touching approximately 0.016% of ...
quantumzeitgeist

Researchers Cut Qc-Mdpc Decoding Failures by 63.7% with Neural Network

Decoding failures that jeopardise next generation encryption are now reduced by 63. 7% thanks to a new neural network ...
eu.36kr

MIT allowed GPT-6 Astra to directly modify Qwen's weights, and its math scores have soared.

MIT has unveiled an innovative, groundbreaking paradigm dubbed "Weight Space Programming", representing a transformative new ...
kucoin

MIT researchers use GPT-6 Astra to enhance Qwen's mathematical ability by 6x

Article by Shenchao TechFlowNo training required, no data needed, no gradient descent.Researchers at MIT fed the weights of ...
BIOENGINEER

Neural network predicts Kerr resonator solitons 20 times faster than simulations

A recurrent neural network trained on simulation data predicts soliton dynamics in Kerr resonators about twenty times faster than traditional Lugiato–Lefever computations.

A CPU running inside an LLM running on a GPU? ...I know it sounds crazy, but here is what I learned and didn't learn from replicating Qwenputer

A model was released on Hugging Face that claims to have a 32-bit RISC-V CPU compiled and embedded within the weights of ...
GitHub

Neural Networks and Deep Learning

B.E. CSE at Osmania University (2014-2018) and M.Tech CS / AI at SSSIHL. - saikrishnavadali05/Academic-CSE-Osmania-SSSIHL ...
BIOENGINEER

Capsule Networks Push Antimicrobial Peptide Prediction Past 97 Percent Accuracy

A capsule neural network model called AMP-CapsNet achieved 97.29 percent test accuracy in classifying antimicrobial peptides ...
Tech Times on MSN

C$1.6 billion Cohere deal lets Ottawa veto any foreign acquisition, no law required

Canada Cohere AI investment: Prime Minister Mark Carney announced a C$1.6 billion equity stake in Toronto-based AI firm ...
news.lavx.hu

LLMs read and edit neural weights without gradients

Astra identifies game policies from raw weight matrices and fixes reasoning loops in Qwen3.5-0.8B by editing 0.016% of parameters, lifting MATH-500 scores from 7.9% to 48.1% without training loops.
GitHub

Carbon-Aware-Deep-Reinforcement-Learning-Platform

1. What exactly is the software product we have created? We have created a production-ready, Kubernetes-native, carbon-aware workload scheduling platform. It operates as a custom Kubernetes controller ...