rikarazome

SWI-Prolog execution for LLMs with CLP(FD), negation-as-failure, and recursion. Benchmarked 90% vs 73% LLM-only accuracy on 30 logic problems.

SWI-Prolog execution for LLMs with CLP(FD) and recursion — boosts logic/constraint accuracy from 73% to 90% on a 30-problem benchmark.