LLM agents that prove every decision and log all work

Original: Provably Auditable and Safe LLM Agents from Human-Authored Ontologies

Source: arxiv.org ↗