Reproduce “Stochastic Dynamics of Generative AI: From Autoregressive Models to Interacting LLM Networks and Neural SDEs” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Vikram Krishnamurthy, Yiming Zhang, Luke Snow
The authors present a tutorial that frames next-token generation, information exchange among language-model agents, and irregularly observed trajectories as learned stochastic systems evolving over time. It connects their distinct analyses without proposing one system for all three. For neural SDE training, it summarizes a conditional cubature approximation rate of O(n−1), versus Monte Carlo’s O(n−1/2), for path-functional objectives under stated assumptions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.