[论文解读] Agentic Framework for Epidemiological Modeling

Rituparna Datta, Zihan Guan+6 more|arXiv (Cornell University)|Jan 30, 2026

COVID-19 epidemiological studies被引用 0

一句话总结

EpiAgent 自动从自然语言场景描述中构建、验证和校准流行病学仿真器，使用流图中介表示和多代理 V&V，提供准确且可解释的反事实投射。

ABSTRACT

Epidemic modeling is essential for public health planning, yet traditional approaches rely on fixed model classes that require manual redesign as pathogens, policies, and scenario assumptions evolve. We introduce EPIAGENT, an agentic framework that automatically synthesizes, calibrates, verifies, and refines epidemiological simulators by modeling disease progression as an iterative program synthesis problem. A central design choice is an explicit epidemiological flow graph intermediate representation that links scenario specifications to model structure and enables strong, modular correctness checks before code is generated. Verified flow graphs are then compiled into mechanistic models supporting interpretable parameter learning under physical and epidemiological constraints. Evaluation on epidemiological scenario case studies demonstrates that EPIAGENT captures complex growth dynamics and produces epidemiologically consistent counterfactual projections across varying vaccination and immune escape assumptions. Our results show that the agentic feedback loop prevents degeneration and significantly accelerates convergence toward valid models by mimicking professional expert workflows.

研究动机与目标

• 在病原体、政策和场景演变时，说明需要可扩展、可适应的流行病建模。
• 引入一个代理化流水线，从场景描述中综合、验证并校准流行病学仿真器。
• 提出流图中介表示，以在代码生成前确保结构正确性和场景保真性。
• demonstrate multi-agent verification and validation to enforce epidemiological validity and interpretability.
• Evaluate EpiAgent on real COVID-19 scenario data showing improved convergence and counterfactual reliability.

提出的方法

• Retrieve-augmented synthesis to generate epidemiological flow graphs from scenario descriptions (RAG with FAISS and sentence-transformer embeddings).
• Iterative graph verification enforcing valid compartments and transitions, plus scenario-consistency checks.
• LLM-based planner instantiates executable simulator code from a verified flow graph using a fixed skeleton and constraint-guided prompts.
• Gradient-based calibration of simulator parameters to observational data with a loss combining level fit and change (MSE on levels and first differences).
• Multi-agent verification and validation modules enforce epidemiological non-negativity, mass conservation, monotonicity, and scenario plausibility, with actionable feedback loops.

Figure 1 : Data flow in EpiAgent has the following structure: Prompt $\;\leftrightarrow\;$ Flow Graph $\;\rightarrow\;$ Simulator Code $\;\rightarrow\;$ Results . The flow graph is an abstract representation of the epidemic model being learned, and supports easier verification. The flow graph is the

实验结果

研究问题

• RQ1 EpiAgent 能否将流行病动态对齐到不同场景下的观测数据？
• RQ2 图验证是否能够阻止结构上无效的模型传播到仿真器？
• RQ3 代码级验证是否能够防止出现低损失但在流行病学上无效的解？
• RQ4 情景条件化模型是否能够产生连贯且可解释的反事实投射？
• RQ5 代理式反馈是否比无引导生成更加加速收敛并提升模型表达力？

主要发现

模型	MAE	MSE	RMSE
State-avg Time-invariant dynamical	3.84e-3	3.10e-5	5.50e-3
State-avg Time-variant (β_t, γ_t)	8.90e-4	1.30e-6	1.10e-3
Neural-incorporated	6.27e-4	7.50e-7	8.00e-4
Nationwide Time-invariant dynamical	2.12e-1	9.57e-2	3.09e-1
Nationwide Time-variant (β_t, γ_t)	4.26e-2	2.53e-3	5.03e-2
Neural-incorporated	2.22e-2	7.13e-4	2.67e-2

• EpiAgent 在各场景上实现了准确拟合，且时变参数提升短期预测准确性。
• 流图验证防止流行病学上无效结构传播到仿真器（未验证的情况下有87%的无效投射）。
• 去除外部知识或流图中介表示会增加误差和结构性问题，凸显该架构的重要性。
• 反事实分析显示在疫苗接种和免疫逃逸假设变化下感染与死亡的连贯转变。
• 代理式反馈循环加速收敛、细化模型表达力，类似专家建模工作流程。
• 该框架可实例化标准行为流行病模型（DDB），并产生经过校准、含不确定性评估的投射。

Figure 2 : Agentic pipeline for epidemic scenario modeling and ensemble simulation. Natural-language scenarios are augmented with domain knowledge (§ 4.1 ), to produce prompts that guide flow-graph synthesis (§ 4.2 ). Generated graphs are iteratively verified to enforce valid compartmental structure