Issue 2026-11-14 · Industry · 研究
Multiverse Computing Tests Quantum Neural Networks in a 30B LLM

Researchers at Multiverse Computing and the University of Toronto demonstrated a multi-head quantum neural network that can replace transformer blocks in a 30B-parameter LLM. A single QNN with 145,281 trainable parameters displaced 16.33% of the model's parameters while maintaining accuracy, and ran on a 156-qubit IBM Heron processor.
Read original ↗- IBM trains small open models to resist prompt injection
- IBM releases LLMON markup language and Granite 4.1 preview models
- IBM and Microsoft launch Imagine Enterprise agentic AI platform
- SAP touts AI tools but discloses no metrics to measure results
- Same model, different harness: task completion rises from 30% to 81%