HawkesLLM: Semantic Uncertainty Propagation in Agentic Text Simulation

In three linesHawkesLLM models semantic uncertainty propagation in multi-agent text simulation systems. A multivariate Hawkes process captures temporal dependencies between text-generating agents, while an LLM generates each new event from compact selected memory. Evaluation on GDELT dataset shows improved late-stage semantic alignment under memory constraints.

Summary generated by Claude — human-verified