Search papers, labs, and topics across Lattice.
3
3
3
7
Unifying diverse mathematical frameworks reveals critical insights into convergence and performance guarantees for reinforcement learning algorithms.
Asynchronous Q-learning converges faster than you thought, with rates up to $n^{-1/6} \log^{4} (nS A)$ now proven for high-dimensional settings.
Ditch reward models: Nash Mirror Prox achieves fast, stable convergence to a Nash equilibrium directly from human preferences, sidestepping the limitations of traditional RLHF.
