01
Multi-agent systems & game theory
Simulation environments for studying equilibrium dynamics, model behavior under asymmetric information, and policy adherence under economic pressure.
Archive
Prior research
Before focusing on STROBE, Dawn State explored learning and failure modes in populations of neural agents. This work is preserved here as earlier research, not a current product line.
02
Adversarial synthetic-data loops
Game-theoretic tests designed to surface unexpected tactics and convert observed failure cases into targeted training data.
03
Weight-space model recombination
Experiments combining parts of neural networks trained in different environments to study whether complementary capabilities can transfer without full retraining.