OntoAct Neuro-Symbolic Framework for Hypothesis-Based Agents

OntoAct utilizes ontological abstraction to convert individual experiences into reusable axioms involving entity types and relations. A large language model proposes hypotheses while a symbolic verifier checks supporting conditions against observations to guide agent actions in environments like Factorio.