Mantis: A Foundation Model for Mechanistic Disease Forecasting

Ananya Sharma; Carson Dudley; Christopher Harding; Emily Martin; Marisa Eisenberg; Reiden Magdaleno

arxiv: 2508.12260 · v5 · submitted 2025-08-17 · 💻 cs.AI · q-bio.QM

Mantis: A Foundation Model for Mechanistic Disease Forecasting

Carson Dudley , Reiden Magdaleno , Christopher Harding , Ananya Sharma , Emily Martin , Marisa Eisenberg This is my paper

Pith reviewed 2026-05-18 22:22 UTC · model grok-4.3

classification 💻 cs.AI q-bio.QM

keywords disease forecastingfoundation modelmechanistic simulationinfectious diseasegeneralizationepidemiologyCOVID-19out-of-distribution

0 comments

The pith

A model trained only on disease simulations outperforms real-data forecasters on COVID-19 and generalizes to many other diseases without seeing any actual records.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper presents Mantis as a foundation model for infectious disease forecasting that is trained exclusively on mechanistic simulations of contagion rather than real epidemiological data. This design aims to overcome the data scarcity, custom training, and expert tuning that limit forecasts for new outbreaks or low-resource settings. Mantis is tested against dozens of existing models across sixteen diseases with varied transmission modes, using metrics for both point accuracy and probabilistic skill. It records lower error than every model in the CDC COVID-19 Forecast Hub on early-pandemic backtests and places in the top two for nearly all other diseases evaluated. The model also succeeds on diseases whose transmission mechanisms were absent from its training simulations, indicating it learns core dynamics instead of memorizing specific patterns.

Core claim

Mantis shows that a single foundation model trained entirely on mechanistic simulations can deliver accurate forecasts for real infectious diseases across regions, outcomes, and transmission types, even when the target disease or mechanism was never present in the training simulations and no real-world data was used at any stage.

What carries the argument

Mantis, a foundation model trained on large-scale mechanistic simulations of contagion dynamics to learn generalizable forecasting behavior.

If this is right

Forecasts become available immediately for new pathogens or regions without first collecting years of case data.
One model can serve many diseases instead of requiring separate tuned models for each.
Performance remains high in settings where historical records are sparse or unreliable.
Models can be updated by expanding the simulation library rather than retraining on new observations.
Forecasts for low-resource areas become feasible without local data collection infrastructure.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same simulation-first approach might extend to forecasting other complex dynamical systems such as economic shocks or ecological invasions.
Hybrid systems could start with Mantis-style pretraining on simulations and then fine-tune lightly on limited real observations when they become available.
If the core dynamics are truly captured, the framework could reduce the need for disease-specific expert knowledge in model construction.

Load-bearing premise

Mechanistic simulations of disease spread contain enough of the real dynamics that a model trained only on them can accurately predict actual historical outbreaks and handle transmission mechanisms it never encountered in training.

What would settle it

Mantis produces higher mean absolute error than standard models when backtested on early data from a future outbreak whose transmission mechanism differs substantially from every simulation family used in its training.

Figures

Figures reproduced from arXiv: 2508.12260 by Ananya Sharma, Carson Dudley, Christopher Harding, Emily Martin, Marisa Eisenberg, Reiden Magdaleno.

**Figure 1.** Figure 1: Conceptual overview of Mantis. Mantis is a simulation-grounded foundation model trained entirely on synthetic outbreaks generated by mechanistic epidemiological models. The training pipeline begins with a modular simulator that encodes diverse outbreak mechanisms, including multiple transmission modes (human-to-human, vectorborne, environmental), progression dynamics, intervention strategies, and populatio… view at source ↗

**Figure 2.** Figure 2: Covariate integration improves accuracy. Mantis maintains calibrated uncertainty across forecast horizons. (a) Including covariates (e.g., using cases to predict hospitalizations) consistently improves Mantis’s accuracy across all forecast horizons. Relative MAE shown for COVID-19 mortality forecasts with (blue) and without (orange) hospitalization covariates across 2, 4, 6, and 8-week horizons. (b) Manti… view at source ↗

**Figure 3.** Figure 3: Mantis Produces Accurate and Generalizable Forecasts Across Diseases and Geographies. (a) Four-week-ahead forecasts (blue dashed line and shaded 90% CI) compared to observed outcomes (black) for COVID-19 mortality in Minnesota and influenza-like illness (ILI) in Michigan. In the latter, Mantis demonstrates its foundation model capacity by accurately forecasting syndromic inputs despite never being traine… view at source ↗

**Figure 4.** Figure 4: Mantis delivers consistent performance across population scales. Relative MAE versus state population for COVID-19 mortality forecasts across 51 U.S. states and territories (Vermont excluded as an outlier). Each point represents the mean relative MAE for a jurisdiction across all forecast dates from April 2020 through November 2021. Population is shown on a logarithmic scale (2020 Census). A weak negative … view at source ↗

read the original abstract

Infectious disease forecasting in novel outbreaks or low-resource settings is hampered by the need for large disease and covariate data sets, bespoke training, and expert tuning, all of which can hinder rapid generation of forecasts for new settings. To help address these challenges, we developed Mantis, a foundation model trained entirely on mechanistic simulations, which enables out-of-the-box forecasting across diseases, regions, and outcomes, even in settings with limited historical data. We evaluated Mantis against 78 forecasting models across sixteen diseases with diverse modes of transmission, assessing both point forecast accuracy (mean absolute error) and probabilistic performance (weighted interval score and coverage). Despite using no real-world data during training, Mantis achieved lower mean absolute error than all models in the CDC's COVID-19 Forecast Hub when backtested on early pandemic forecasts which it had not previously seen. Across all other diseases tested, Mantis consistently ranked in the top two models across evaluation metrics. Mantis further generalized to diseases with transmission mechanisms not represented in its training data, demonstrating that it can capture fundamental contagion dynamics rather than memorizing disease-specific patterns. These capabilities illustrate that purely simulation-based foundation models such as Mantis can provide a practical foundation for disease forecasting: general-purpose, accurate, and deployable where traditional models struggle.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Mantis trains a foundation model only on mechanistic simulations and claims it beats data-driven forecasters on real outbreaks including early COVID, but the results rest on how well those simulations cover real reporting and behavior effects.

read the letter

The main thing to know is that this paper trains a foundation model called Mantis entirely on mechanistic simulations and claims it can forecast real disease outbreaks better than many existing models without using any real data in training. It reports lower mean absolute error than all CDC COVID hub models on early pandemic backtests and top-two rankings across 16 diseases with different transmission modes, plus generalization to mechanisms absent from training. That combination is new relative to the data-driven or hybrid methods referenced in the abstract. Most prior forecasting work either tunes directly on historical series or mixes limited real data with simulations. Here the model learns solely from simulated trajectories and then applies zero-shot to real cases, which is a distinct setup if the numbers hold. The paper does a good job laying out a broad evaluation: comparisons to 78 models, both point and probabilistic metrics, and explicit tests on unseen mechanisms. If the backtest protocol and model implementations are clean, this points to a practical tool for low-data or novel-outbreak settings. The central soft spot is whether the simulations embed enough real-world nuisance factors. Standard compartmental models often simplify reporting delays, testing rates, mobility changes, and superspreading. For the claimed MAE advantage and cross-mechanism generalization to be robust, the training ensemble needs to vary those parameters at scale. The abstract gives no quantitative detail on the simulator ranges or coverage of those effects, so it remains possible the model is picking up simulation artifacts rather than transferable dynamics. That concern is real but not fatal on current evidence; it is the main place where extra checks would help. This paper is aimed at researchers building general forecasting systems for public health, especially those interested in simulation-based machine learning or rapid deployment in data-scarce regions. A reader working on epi modeling or foundation models for scientific domains would get concrete value from the setup and results. The work shows clear engagement with the forecasting literature and honest framing of its assumptions. It deserves a serious referee to examine the methods, parameter choices, and statistical comparisons. I recommend sending it to peer review rather than desk rejecting it.

Referee Report

2 major / 2 minor

Summary. The manuscript introduces Mantis, a foundation model for infectious disease forecasting trained exclusively on mechanistic simulations without any real-world data. It evaluates the model against 78 forecasting models across 16 diseases with diverse transmission modes, claiming lower mean absolute error than all CDC COVID-19 Forecast Hub models on early-pandemic backtests, top-two rankings on other diseases for MAE, weighted interval score, and coverage, and generalization to transmission mechanisms absent from the training simulations.

Significance. If the results hold, the work has substantial significance for mechanistic and AI-driven epidemiology. A simulation-only foundation model that delivers zero-shot superiority on real historical data and generalizes across unseen transmission modes would address a core limitation of data-intensive or disease-specific forecasters, enabling rapid deployment in novel outbreaks or low-resource settings. The approach of learning transferable contagion dynamics from simulations rather than empirical fitting is a clear strength and could shift the field toward more general-purpose tools.

major comments (2)

[Methods] Simulation design (Methods section): The central generalization claim—that Mantis captures fundamental dynamics rather than simulation-specific artifacts—requires explicit evidence that the mechanistic simulator ensemble spans real-world nuisance parameters at sufficient scale. The abstract asserts zero-shot MAE superiority and cross-mechanism generalization, yet no quantitative coverage analysis is provided for reporting delays, testing rates, mobility changes, or superspreading heterogeneity; if these factors are under-represented or fixed in the training distribution, the performance gap on real data may not transfer as claimed.
[Results] COVID-19 backtest evaluation (Results section): The headline result that Mantis outperforms every model in the CDC Forecast Hub on early-pandemic forecasts is load-bearing for the zero-shot claim. The manuscript must specify the exact backtest periods, the precise subset of hub models included in the comparison, data exclusion rules, and any statistical significance tests; without these details, it remains possible that evaluation choices (e.g., period selection or model filtering) drive the reported MAE advantage.

minor comments (2)

[Evaluation Metrics] Clarify the exact definitions and computation of the weighted interval score and coverage metrics in the evaluation protocol to support reproducibility.
[Figures] Ensure all performance comparison figures include clear axis labels, legends, and error bars or confidence intervals for the reported rankings.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments and positive assessment of the significance of our work. We address each major comment below and outline revisions that will strengthen the manuscript's clarity and rigor.

read point-by-point responses

Referee: [Methods] Simulation design (Methods section): The central generalization claim—that Mantis captures fundamental dynamics rather than simulation-specific artifacts—requires explicit evidence that the mechanistic simulator ensemble spans real-world nuisance parameters at sufficient scale. The abstract asserts zero-shot MAE superiority and cross-mechanism generalization, yet no quantitative coverage analysis is provided for reporting delays, testing rates, mobility changes, or superspreading heterogeneity; if these factors are under-represented or fixed in the training distribution, the performance gap on real data may not transfer as claimed.

Authors: We agree that a quantitative coverage analysis would provide stronger support for the generalization claims. In the revised manuscript, we will add a new subsection to the Methods that reports the parameter ranges and sampling distributions used in the simulation ensemble for reporting delays, testing rates, mobility changes, and superspreading heterogeneity. We will include summary statistics and visualizations comparing these ranges to empirical values drawn from the literature for the diseases under study, thereby demonstrating that the training distribution is sufficiently broad to support zero-shot transfer and cross-mechanism generalization. revision: yes
Referee: [Results] COVID-19 backtest evaluation (Results section): The headline result that Mantis outperforms every model in the CDC Forecast Hub on early-pandemic forecasts is load-bearing for the zero-shot claim. The manuscript must specify the exact backtest periods, the precise subset of hub models included in the comparison, data exclusion rules, and any statistical significance tests; without these details, it remains possible that evaluation choices (e.g., period selection or model filtering) drive the reported MAE advantage.

Authors: We appreciate the need for full transparency on the evaluation protocol. The revised Results section will explicitly list: (i) the precise calendar periods used for the early-pandemic backtests, (ii) the complete set of CDC Forecast Hub models included in the comparison along with inclusion criteria, (iii) any data exclusion rules (e.g., handling of missing observations or specific locations), and (iv) statistical significance tests (paired Wilcoxon or t-tests on MAE differences with p-values) confirming that the reported advantage is robust to evaluation choices. These details will be added without altering the underlying experimental outcomes. revision: yes

Circularity Check

0 steps flagged

No circularity: training on simulations with independent real-world evaluation

full rationale

The paper trains Mantis exclusively on mechanistic simulations and evaluates point and probabilistic forecasts on separate real-world data (e.g., CDC COVID-19 Forecast Hub backtests and other disease datasets) that were never seen during training. The abstract and description present these MAE, WIS, and coverage results as empirical outcomes against external models rather than quantities obtained by fitting parameters to the target evaluation data or by self-referential definitions. No equations, self-citations, or ansatzes are shown that would reduce the claimed generalization or performance to inputs by construction. The derivation chain therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on the assumption that mechanistic simulations encode transferable contagion dynamics; no free parameters or invented entities are described in the abstract.

axioms (1)

domain assumption Mechanistic simulations capture fundamental contagion dynamics sufficiently for generalization to real data and unseen diseases.
Invoked to support out-of-the-box performance and generalization claims.

pith-pipeline@v0.9.0 · 5767 in / 1303 out tokens · 52052 ms · 2026-05-18T22:22:59.748914+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Foundation/RealityFromDistinction.lean reality_from_one_distinction unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

Mantis was trained entirely on synthetic outbreaks generated using mechanistic epidemiological models... SEAIR structure... stochastic compartmental model... human-to-human, vector-borne, environmental transmission
IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

trained on over 400 million simulated days... no real-world data during training... generalized to diseases with transmission mechanisms not represented in its training data

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Forward citations

Cited by 3 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

Latent Chain-of-Thought Improves Structured-Data Transformers
cs.LG 2026-05 unverdicted novelty 6.0

Latent chain-of-thought via recurrent feedback tokens improves average performance of structured-data transformers on time-series forecasting and tabular prediction.
In-Context Learning Under Regime Change
cs.LG 2026-04 unverdicted novelty 6.0

Transformers can solve in-context change-point detection with model size scaling by knowledge of the shift timing, matching optimal baselines on synthetic data and improving pretrained models on disease and financial ...
Prediction Markets Underperform Simple Baselines For Infectious Disease Forecasting
stat.AP 2026-05 conditional novelty 4.0

Prediction markets fail to outperform standard benchmarks for forecasting influenza hospitalizations and measles cases.

Reference graph

Works this paper leans on

22 extracted references · 22 canonical work pages · cited by 3 Pith papers · 1 internal anchor

[1]

Role of modelling in covid-19 policy development.Paediatric Respiratory Reviews, 35:57–60, Sep 2020

Emma S McBryde, Michael T Meehan, Oyelola A Adegboye, Adeshina I Adekunle, Jamie M Caldwell, Anton Pak, Diana P Rojas, Bridget M Williams, and James M Trauer. Role of modelling in covid-19 policy development.Paediatric Respiratory Reviews, 35:57–60, Sep 2020

work page 2020
[2]

Cramer, Evan L

Estee Y. Cramer, Evan L. Ray, Velma K. Lopez, Johannes Bracher, Andrea Brennen, Alvaro J. Castro Rivadeneira, Aaron Gerding, Tilmann Gneiting, Katie H. House, Yuxin Huang, Dasuni Jayawardena, Abdul H. Kanji, Ayush Khandelwal, Khoa Le, Anja M¨ uhlemann, Jarad Niemi, Apurv Shah, Ariane Stark, Yijin Wang, Nutcha Wattanachit, Martha W. Zorn, Youyang Gu, Sansi...

work page 2022
[3]

Deepcovid: An operational deep learning-driven framework for explain- able real-time covid-19 forecasting

Alexander Rodriguez, Anika Tabassum, Jiaming Cui, Jiajia Xie, Javen Ho, Pulak Agarwal, Bijaya Adhikari, and Aditya Prakash. Deepcovid: An operational deep learning-driven framework for explain- able real-time covid-19 forecasting. InProceedings of the AAAI Conference on Artificial Intelligence, volume 35, 2021

work page 2021
[4]

Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting, 2021

Dongxia Wu, Liyao Gao, Xinyue Xiong, Matteo Chinazzi, Alessandro Vespignani, Yi-An Ma, and Rose Yu. Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting, 2021

work page 2021
[5]

Sebastian Funk, Anton Camacho, Adam J Kucharski, Rachel Lowe, Rosalind M Eggo, and W John Edmunds. Assessing the performance of real-time epidemic forecasts: A case study of ebola in the western area region of sierra leone, 2014–15.PLOS Computational Biology, 15(2):e1006785, 2019

work page 2014
[6]

Reich et al

Nicholas G. Reich et al. Collaborative hubs: Making the most of predictive epidemic modeling.American Journal of Public Health, 112(6):839–842, 2022. Epub 2022 Apr 14

work page 2022
[7]

Runge et al

Michael C. Runge et al. Scenario design for infectious disease projections: Integrating concepts from decision analysis and experimental design.Epidemics, 47:100775, 2024. Epub 2024 May 24

work page 2024
[8]

Simulation as Supervision: Mechanistic Pretraining for Scientific Discovery

Carson Dudley, Reiden Magdaleno, Christopher Harding, and Marisa Eisenberg. Simulation as super- vision: Mechanistic pretraining for scientific discovery.arXiv preprint arXiv:2507.08977, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025
[9]

Defsi: Deep learning based epidemic forecasting with synthetic information

Lijing Wang, Jiangzhuo Chen, and Madhav Marathe. Defsi: Deep learning based epidemic forecasting with synthetic information. InProceedings of the AAAI Conference on Artificial Intelligence, volume 33, 2019

work page 2019
[10]

Mi safe start map

University of Michigan. Mi safe start map. Accessed August 15, 2025

work page 2025
[11]

Farrow, Logan C

David C. Farrow, Logan C. Brooks, Aaron Rumack, Ryan J. Tibshirani, and Roni Rosenfeld. Delphi epidata api.https://github.com/cmu-delphi/delphi-epidata, 2015. Carnegie Mellon University, Delphi Research Group

work page 2015
[12]

Weekly united states covid-19 cases and deaths by state - archived.https: //data.cdc.gov/Case-Surveillance/Weekly-United-States-COVID-19-Cases-and-Deaths-by-/ pwn4-m3yp, 2025

CDC COVID-19 Response. Weekly united states covid-19 cases and deaths by state - archived.https: //data.cdc.gov/Case-Surveillance/Weekly-United-States-COVID-19-Cases-and-Deaths-by-/ pwn4-m3yp, 2025. 11

work page 2025
[13]

Counts of dengue without warning signs reported in brazil: 1980–2005 (2.0) [data set]

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of dengue without warning signs reported in brazil: 1980–2005 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/BR.722862003,

work page doi:10.25337/t7/ptycho.v2.0/br.722862003 1980
[14]

Counts of viral hepatitis type b reported in united states of america: 1951–2007 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/ US.66071002, 2018

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of viral hepatitis type b reported in united states of america: 1951–2007 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/ US.66071002, 2018. Project Tycho

work page doi:10.25337/t7/ptycho.v2.0/ 1951
[15]

Counts of smallpox reported in united states of america: 1888–1952 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.67924001,

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of smallpox reported in united states of america: 1888–1952 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.67924001,

work page doi:10.25337/t7/ptycho.v2.0/us.67924001 1952
[16]

Counts of scarlet fever reported in united states of america: 1888–1969 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.30242009,

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of scarlet fever reported in united states of america: 1888–1969 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.30242009,

work page doi:10.25337/t7/ptycho.v2.0/us.30242009 1969
[17]

Ray, Tilmann Gneiting, and Nicholas G

Johannes Bracher, Evan L. Ray, Tilmann Gneiting, and Nicholas G. Reich. Evaluating epidemic forecasts in an interval format.PLOS Computational Biology, 17(2):e1008618, 2021

work page 2021
[18]

Maddix, Hao Wang, Michael W

Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Sundar Rangapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschiegner, Danielle C. Maddix, Hao Wang, Michael W. Mahoney, Kari Torkkola, Andrew Gordon Wilson, Michael Bohlke-Schneider, and Yuyang Wang. Chronos: Transformer-based lang...

work page 2024
[19]

Foundation time series models for forecasting and policy evaluation in infectious disease epidemics.medRxiv, February 2025

Suprabhath Kalahasti, Benjamin Faucher, Boxuan Wang, Claudio Ascione, Ricardo Carbajal, Maxime Enault, Christophe Vincent Cassis, Titouan Launay, Caroline Guerrisi, Pierre-Yves Bo¨ elle, Federico Baldo, and Eugenio Valdano. Foundation time series models for forecasting and policy evaluation in infectious disease epidemics.medRxiv, February 2025. preprint,...

work page 2025
[20]

Decoupled weight decay regularization, 2019

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization, 2019

work page 2019
[21]

Estimating the latent period of coronavirus disease 2019 (covid-19)

Hualei Xin, Yu Li, Peng Wu, Zhili Li, Eric HY Lau, Ying Qin, Liping Wang, Benjamin J Cowling, Tim K Tsang, and Zhongjie Li. Estimating the latent period of coronavirus disease 2019 (covid-19). Clinical Infectious Diseases, 74(9):1678–1681, 2022

work page 2019
[22]

Vincent Ka Chun Yan, Eric Yuk Fai Wan, Xuxiao Ye, Anna Hoi Ying Mok, Francisco Tsz Tsun Lai, Celine Sze Ling Chui, Xue Li, Carlos King Ho Wong, Philip Hei Li, Tiantian Ma, Simon Qin, Chak Sing Lau, Ian Chi Kei Wong, and Esther Wai Yin Chan. Waning effectiveness against covid-19-related hospitalization, severe complications, and mortality with two to three...

work page 2023

[1] [1]

Role of modelling in covid-19 policy development.Paediatric Respiratory Reviews, 35:57–60, Sep 2020

Emma S McBryde, Michael T Meehan, Oyelola A Adegboye, Adeshina I Adekunle, Jamie M Caldwell, Anton Pak, Diana P Rojas, Bridget M Williams, and James M Trauer. Role of modelling in covid-19 policy development.Paediatric Respiratory Reviews, 35:57–60, Sep 2020

work page 2020

[2] [2]

Cramer, Evan L

Estee Y. Cramer, Evan L. Ray, Velma K. Lopez, Johannes Bracher, Andrea Brennen, Alvaro J. Castro Rivadeneira, Aaron Gerding, Tilmann Gneiting, Katie H. House, Yuxin Huang, Dasuni Jayawardena, Abdul H. Kanji, Ayush Khandelwal, Khoa Le, Anja M¨ uhlemann, Jarad Niemi, Apurv Shah, Ariane Stark, Yijin Wang, Nutcha Wattanachit, Martha W. Zorn, Youyang Gu, Sansi...

work page 2022

[3] [3]

Deepcovid: An operational deep learning-driven framework for explain- able real-time covid-19 forecasting

Alexander Rodriguez, Anika Tabassum, Jiaming Cui, Jiajia Xie, Javen Ho, Pulak Agarwal, Bijaya Adhikari, and Aditya Prakash. Deepcovid: An operational deep learning-driven framework for explain- able real-time covid-19 forecasting. InProceedings of the AAAI Conference on Artificial Intelligence, volume 35, 2021

work page 2021

[4] [4]

Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting, 2021

Dongxia Wu, Liyao Gao, Xinyue Xiong, Matteo Chinazzi, Alessandro Vespignani, Yi-An Ma, and Rose Yu. Deepgleam: A hybrid mechanistic and deep learning model for covid-19 forecasting, 2021

work page 2021

[5] [5]

Sebastian Funk, Anton Camacho, Adam J Kucharski, Rachel Lowe, Rosalind M Eggo, and W John Edmunds. Assessing the performance of real-time epidemic forecasts: A case study of ebola in the western area region of sierra leone, 2014–15.PLOS Computational Biology, 15(2):e1006785, 2019

work page 2014

[6] [6]

Reich et al

Nicholas G. Reich et al. Collaborative hubs: Making the most of predictive epidemic modeling.American Journal of Public Health, 112(6):839–842, 2022. Epub 2022 Apr 14

work page 2022

[7] [7]

Runge et al

Michael C. Runge et al. Scenario design for infectious disease projections: Integrating concepts from decision analysis and experimental design.Epidemics, 47:100775, 2024. Epub 2024 May 24

work page 2024

[8] [8]

Simulation as Supervision: Mechanistic Pretraining for Scientific Discovery

Carson Dudley, Reiden Magdaleno, Christopher Harding, and Marisa Eisenberg. Simulation as super- vision: Mechanistic pretraining for scientific discovery.arXiv preprint arXiv:2507.08977, 2025

work page internal anchor Pith review Pith/arXiv arXiv 2025

[9] [9]

Defsi: Deep learning based epidemic forecasting with synthetic information

Lijing Wang, Jiangzhuo Chen, and Madhav Marathe. Defsi: Deep learning based epidemic forecasting with synthetic information. InProceedings of the AAAI Conference on Artificial Intelligence, volume 33, 2019

work page 2019

[10] [10]

Mi safe start map

University of Michigan. Mi safe start map. Accessed August 15, 2025

work page 2025

[11] [11]

Farrow, Logan C

David C. Farrow, Logan C. Brooks, Aaron Rumack, Ryan J. Tibshirani, and Roni Rosenfeld. Delphi epidata api.https://github.com/cmu-delphi/delphi-epidata, 2015. Carnegie Mellon University, Delphi Research Group

work page 2015

[12] [12]

Weekly united states covid-19 cases and deaths by state - archived.https: //data.cdc.gov/Case-Surveillance/Weekly-United-States-COVID-19-Cases-and-Deaths-by-/ pwn4-m3yp, 2025

CDC COVID-19 Response. Weekly united states covid-19 cases and deaths by state - archived.https: //data.cdc.gov/Case-Surveillance/Weekly-United-States-COVID-19-Cases-and-Deaths-by-/ pwn4-m3yp, 2025. 11

work page 2025

[13] [13]

Counts of dengue without warning signs reported in brazil: 1980–2005 (2.0) [data set]

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of dengue without warning signs reported in brazil: 1980–2005 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/BR.722862003,

work page doi:10.25337/t7/ptycho.v2.0/br.722862003 1980

[14] [14]

Counts of viral hepatitis type b reported in united states of america: 1951–2007 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/ US.66071002, 2018

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of viral hepatitis type b reported in united states of america: 1951–2007 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/ US.66071002, 2018. Project Tycho

work page doi:10.25337/t7/ptycho.v2.0/ 1951

[15] [15]

Counts of smallpox reported in united states of america: 1888–1952 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.67924001,

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of smallpox reported in united states of america: 1888–1952 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.67924001,

work page doi:10.25337/t7/ptycho.v2.0/us.67924001 1952

[16] [16]

Counts of scarlet fever reported in united states of america: 1888–1969 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.30242009,

Willem Van Panhuis, Abigail Cross, and Donald Burke. Counts of scarlet fever reported in united states of america: 1888–1969 (2.0) [data set].https://doi.org/10.25337/T7/ptycho.v2.0/US.30242009,

work page doi:10.25337/t7/ptycho.v2.0/us.30242009 1969

[17] [17]

Ray, Tilmann Gneiting, and Nicholas G

Johannes Bracher, Evan L. Ray, Tilmann Gneiting, and Nicholas G. Reich. Evaluating epidemic forecasts in an interval format.PLOS Computational Biology, 17(2):e1008618, 2021

work page 2021

[18] [18]

Maddix, Hao Wang, Michael W

Abdul Fatir Ansari, Lorenzo Stella, Caner Turkmen, Xiyuan Zhang, Pedro Mercado, Huibin Shen, Oleksandr Shchur, Syama Sundar Rangapuram, Sebastian Pineda Arango, Shubham Kapoor, Jasper Zschiegner, Danielle C. Maddix, Hao Wang, Michael W. Mahoney, Kari Torkkola, Andrew Gordon Wilson, Michael Bohlke-Schneider, and Yuyang Wang. Chronos: Transformer-based lang...

work page 2024

[19] [19]

Foundation time series models for forecasting and policy evaluation in infectious disease epidemics.medRxiv, February 2025

Suprabhath Kalahasti, Benjamin Faucher, Boxuan Wang, Claudio Ascione, Ricardo Carbajal, Maxime Enault, Christophe Vincent Cassis, Titouan Launay, Caroline Guerrisi, Pierre-Yves Bo¨ elle, Federico Baldo, and Eugenio Valdano. Foundation time series models for forecasting and policy evaluation in infectious disease epidemics.medRxiv, February 2025. preprint,...

work page 2025

[20] [20]

Decoupled weight decay regularization, 2019

Ilya Loshchilov and Frank Hutter. Decoupled weight decay regularization, 2019

work page 2019

[21] [21]

Estimating the latent period of coronavirus disease 2019 (covid-19)

Hualei Xin, Yu Li, Peng Wu, Zhili Li, Eric HY Lau, Ying Qin, Liping Wang, Benjamin J Cowling, Tim K Tsang, and Zhongjie Li. Estimating the latent period of coronavirus disease 2019 (covid-19). Clinical Infectious Diseases, 74(9):1678–1681, 2022

work page 2019

[22] [22]

Vincent Ka Chun Yan, Eric Yuk Fai Wan, Xuxiao Ye, Anna Hoi Ying Mok, Francisco Tsz Tsun Lai, Celine Sze Ling Chui, Xue Li, Carlos King Ho Wong, Philip Hei Li, Tiantian Ma, Simon Qin, Chak Sing Lau, Ian Chi Kei Wong, and Esther Wai Yin Chan. Waning effectiveness against covid-19-related hospitalization, severe complications, and mortality with two to three...

work page 2023