Research
What I'm chasing
Four directions right now, and they overlap: agents, the harnesses around them, and how much you can get out of a model without making it bigger. I'm in the MASTERS group (MultiAgent SysTEms ReSearch) at the University of Tulsa, advised by Sandip Sen.
Most of it is active work. The published piece is below: a workshop paper on when an agent should disobey.
The directions
- Agent harnesses
- What you can get out of agents through orchestration and harness design. This is where the PhD is converging; Synapse and the daily practice are the lab bench.
- Multi-agent negotiation
- How LLM agents negotiate and coordinate, including when principled disobedience helps the human. The framework and its domains are on GitHub.
- Small models, pushed further
- Fine-tuning and orchestration to get more out of small, local models, so we lean on the big ones less. Ongoing unpublished work: continuous negotiation-stance control for a 7B model from one trained adapter, no retraining.
- The space lane
- LODA, an architecture for objective-informed deviation in LLM agents, framed for rovers. Presented at NASA JPL during a MASTERS group visit.
The paper
Published work
Disobedience for a Cause: Leveraging Implicit Objectives in User Plans by LLM Surrogates
With Dale Peasley and Sandip Sen. It asks when an agent should override your literal instructions because it understands what you actually want. I presented it at the workshop in Detroit.
sit at the desk →(the 3D version)There's a 3D version of this desk. It needs a bigger screen.This browser doesn't run WebGL.It moves, and your system asks for less motion.