Quantum physicist have shrunk and de-censored deep seek R1

Quantum physicists have developed a method to compress and de-censor large language models, specifically the DeepSeek R1, using a quantum physics-inspired approach. This technique, employed by Multiverse, involves tensor networks to efficiently represent and manipulate data, allowing for the precise removal of specific information. The method was tested on questions known to be restricted in Chinese models, with the modified model providing factual responses similar to Western models. This approach aims to make AI models more efficient while maintaining performance, addressing the challenge of compressing large models without losing capability. The research also highlights the broader issue of censorship in AI models, particularly those originating from China, and the ongoing efforts to mitigate these restrictions.