PapersBuilders10:42Chinese contentHSI框架:让冻结LLM通过分层自改进进化任务脚手架想用冻结模型提升agent性能?HSI让模型自己改写执行脚手架,在BabyAI等任务上涨了39个点,代码已开源,值得一看。#HSI#DeepSeek-V4-Flash-Preview#BALROG#agentsaarXiv: DeepSeek@Tailin ZhouSourceRead laterReadWorth followingUsefulFollow HSI
Papers09:18Chinese content分层Solomonoff归纳:一种无界机器学习模型这篇把Solomonoff归纳改成能基于数据集预测,证明与原版等价且误差有界,做理论的人可以看。#Solomonoff Induction#HSI#machine learning theoryaarXiv cs.AI@Nathan YoungSourceRead laterReadWorth followingUsefulFollow Solomonoff Induction