Research
The foundation of the work
A sample of our foundational research. It is the evidence that supports the work and reflects that we measure before we claim. We share the principles and the questions we pursue; the experimental setups and detailed results stay private.
-
Agreement among agents without a central coordinator
How several agents converge on a shared goal and divide the work with no single point of control.
We study agreement protocols that let a group of agents coordinate stably, even when each one knows only part of the problem. Our interest is what holds that coordination together as the number of agents grows or conditions shift, and how to keep a single failure from pulling down the whole. We share the principles and the questions; the experimental setups stay private.
-
Self-verification before acting
Methods for a system to assess the quality of its own output and decide whether to proceed, correct, or stop.
We work on internal confidence signals a system can measure over its own work, so it knows when to proceed, when to correct, and when to ask a person for help. The underlying question is how to calibrate that confidence so it is neither reckless nor paralyzing.
-
Decisions anchored in evidence
How to keep a clear trace of why an autonomous system made each decision.
We research ways for an autonomous decision to stay linked to the evidence that supports it, so a person can reconstruct the reasoning afterward. We care that the trace stays clear and proportionate when a decision chains many steps together.