NeSy-RAG compiles retrieved chunks into executable Prolog and lifts ShARC from 42.8% to 61.1%.

Instead of feeding retrieved text to an LLM and hoping the reasoning holds, arXiv 2608.06292 synthesizes a Prolog module per chunk, generating predicates encoding Boolean claims that may depend on user-specific facts, then retrieves and composes them into queries using joint natural-language/code embeddings. A symbolic knowledge-gap detector identifies missing user facts whose truth value would change the outcome and automatically triggers a follow-up question. That's a real answer to the "agent guesses instead of asking" problem, and it's decidable rather than vibes-based. 61.1% versus 42.8% for a same-model RAG baseline, no domain-specific training, and every answer ships an execution trace linking each step to its source.