| Layer | Powered by | Strength |
|---|---|---|
| Neural | LLM + vector embeddings | Natural-language understanding |
| Symbolic | RDF graph + SPARQL engine | Deterministic logic and precision |
Why both?
Neural networks excel at interpreting meaning but can hallucinate. Symbolic systems are precise but brittle with unstructured input. By fusing the two, Worlds lets an agent understand a question via the neural layer, and prove the answer via the symbolic layer.In practice
- Ingestion — An LLM extracts structured triples from raw text.
- Retrieval — Hybrid search combines vector similarity with graph traversal.
- Reasoning — SPARQL queries return verifiable facts the agent can cite.
Learn more
- Philosophy — principles behind the approach
- Architecture — system-level deep dive