Our Origin
Oraculab was born in late 2023 from a series of late-night conversations between three researchers — a computational linguist, a knowledge graph specialist, and a reinforcement learning engineer. They shared a growing concern: as language models grew larger, their understanding of the world wasn't growing deeper. Scaling laws were delivering fluency, not comprehension.
They founded Oraculab to pursue a different path — one where neural networks and structured knowledge reinforce each other, where agents can plan and reason explicitly, and where AI systems can explain their reasoning in terms humans can verify. The name "Oraculab" combines "oracle" (knowledge) with "lab" (experimentation) — reflecting our commitment to systematic exploration of intelligence.
Research Philosophy
We believe the next leap in AI won't come from bigger models alone. It will come from better architectures that combine the pattern-matching power of neural networks with the precision of symbolic reasoning. Our research spans knowledge representation, agentic planning, mechanistic interpretability, and multi-modal understanding — all unified by the goal of building AI that genuinely knows things, not just predicts tokens.
We are committed to open science. Every paper we publish includes reproducible code, datasets, and model weights. We believe transparency accelerates progress and builds trust.
Our Principles
🔍 Rigorous Inquiry
Hypothesis-driven research with empirical validation. We don't chase hype; we chase understanding.
🌍 Global Collaboration
Our team spans 9 countries. Diverse perspectives produce better science.
📖 Open by Default
Papers, code, data, models — all open. Knowledge locked behind paywalls helps no one.
⚖️ Responsible AI
Alignment and safety aren't afterthoughts. They're core research priorities from day one.