Runtime Uncertainty Monitoring for LLM-Based Multi-Agent Systems Using Bayesian Networks
Bart Custers, Koorosh Aslansefat
Why It Matters
What makes this one worth your time
Understanding and managing uncertainty in LLM-based systems is crucial for high-stakes applications like actuarial risk modeling, where errors can have significant consequences.
A framework for runtime uncertainty monitoring in LLM-based multi-agent systems using Bayesian Networks.
Summary
The paper proposes a framework for monitoring runtime uncertainty in multi-agent systems that utilize large language models, specifically for actuarial risk modeling. It introduces a method for uncertainty propagation using token-level log-probabilities and Bayesian Networks, transforming these into calibrated task-level confidence estimates.
Key contributions
- A novel approach to uncertainty propagation using token-level log-probabilities.
- Integration of Bayesian Networks for runtime uncertainty monitoring in multi-agent systems.
Notable insights
- The use of token-level log-probabilities transformed into calibrated confidence estimates is a novel approach to uncertainty quantification.
- Incorporating Bayesian Networks for uncertainty propagation in multi-agent systems is a clever integration of probabilistic reasoning.
Possible limitations
- Not stated in the abstract
Abstract
arXiv:2607.25877v1 Announce Type: new Abstract: This paper investigates how multi-agent systems (MAS)-based on large language models (LLMs) can support actuarial risk modelling, with a particular focus on uncertainty quantification. Actuarial workflows represent a high-stakes decision-support setting where unreliable outputs may lead to incorrect risk assessment, unfair pricing, and regulatory non-compliance. To address uncertainty introduced by the probabilistic nature of LLMs and dependencies between agents, a multi-agent framework is proposed in which specialised agents perform data preparation, modelling, review, and explanation tasks under a central hub. The main contribution is a novel approach to uncertainty propagation using token-level log-probabilities and a Bayesian Network. Importantly, log probabilities are not treated as direct probabilities of correctness or task success. Instead, length-normalised log-probability summaries are transformed into calibrated task-level confidence estimates before incorporation into the Bayesian Network. Results show that the framework reproduces baseline actuarial performance while providing additional insight into workflow stability and runtime uncertainty propagation.