04:00
2026-08-04
machinebrief.com
large-language-models
Cloud-ScPO: Hidden-State Geometry for Semi-Supervised Preference Optimization in LLM Reasoning
A new arXiv preprint (2608.01014v1) introduces Cloud-ScPO, a topology-guided preference-mining framework that derives preference supervision from the hidden-state geometry of large language models (LL…