Search papers, labs, and topics across Lattice.
Affiliation:
2
0
3
This work reduces state-adversarial Markov decision processes to a strategically equivalent constrained zero-sum one-sided partially observable stochastic game and presents the first algorithmic route to computing $\epsilon$-approximations of initial-state dependent equilibria.
LLM agents' internal activations leak detectable signals of collusion, even when transferred to structurally different multi-agent scenarios, opening a new frontier for AI safety.
