Tau Intelligence is a research group. The name τ stands for Trustworthy Autonomy under Uncertainty — the property we want every system we touch to have.
Our work sits at the intersection of reinforcement learning, learned world models, and robotics. We care about the unglamorous parts: stable training, honest evaluation, robust deployment, and methods that survive contact with reality.
We publish openly at venues like ICML, TMLR, RLC, ICLR, AAAI, etc., and we open-source what we can.
Strong baselines, ablations, and confidence intervals — not just headline numbers.
We default to open-sourcing what we can and writing things down so others can build on them.
Models that know what they don't know are the only ones safe to deploy.
We pick problems whose solutions matter five years from now, not five weeks.
For research collaborations, questions about our work, or academic correspondence — the channels below all reach us.
