HawkesLLM | Zewei Deng

HawkesLLM combines multivariate Hawkes processes with language models to model temporal influence and semantic uncertainty propagation in agentic text simulation.

The framework computes Hawkes influence scores over the event history, selects the most influential predecessor events under a constrained prompt-memory budget, and uses that compact context to generate the next event.

Hawkes influence scores select the most relevant predecessor events, which are assembled into a compact prompt for the next-event generation step.

(Deng et al., 2026)

References

2026

HawkesLLM: Semantic Uncertainty Propagation in Agentic Text Simulation

Zewei Deng, Tinghan Ye, and Liyan Xie

In ICML Workshop on Statistical Frameworks for Uncertainty in Agentic Systems, 2026

arXiv Bib

@inproceedings{deng2026hawkesllm,
  title = {{HawkesLLM}: Semantic Uncertainty Propagation in Agentic Text Simulation},
  author = {Deng, Zewei and Ye, Tinghan and Xie, Liyan},
  booktitle = {ICML Workshop on Statistical Frameworks for Uncertainty in Agentic Systems},
  year = {2026},
}