Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
A Counterfactual Semantics for Hybrid Dynamical Systems
Authors: Andy Zane, Dmitry Batenkov, Rafal Urbaniak, Jeremy Zucker, Sam Witty
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Theoretical | Our contributions are: 1. A formal, counterfactual semantics for dynamically triggered, instantaneous interventions in continuous-time dynamical systems. 2. Under minimal requirements on interventional specifications, proof that sufficient conditions for solution existence, uniqueness, and finite-time measurability are preserved in the intervened system. Our framework also explicitly connects to established well-posedness conditions on hybrid dynamical systems. 3. A case study on the probabilities of necessary and sufficient causation applied to fishery management, demonstrating extensibility to non-trivial causal estimands rarely applied to dynamical systems. Neur IPS Paper Checklist...4. Experimental result reproducibility...Answer: [NA] Justification: We do not include experiments in the main body, but do provide some simulated estimation results in the appendix. We provide ample model details therein for reproducibility and provide links to our simulation code. |
| Researcher Affiliation | Collaboration | Andy Zane1,2 EMAIL Dmitry Batenkov1 EMAIL Rafal Urbaniak1 EMAIL Jeremy Zucker3 EMAIL Sam Witty4 EMAIL 1Basis Research Institute, New York, NY 2University of Massachusetts Amherst, Amherst, MA 3Pacific Northwest National Laboratory, Richland, WA 4Sorbus AI |
| Pseudocode | No | The paper describes methods using mathematical definitions, propositions, lemmas, and theorems. There are no explicitly labeled "Pseudocode" or "Algorithm" blocks present in the document. |
| Open Source Code | Yes | Code is available here,13 and relies on the dynamical systems package from the causal probabilistic programming language Chi Rho (Basis-Research, 2025).14 13https://basisresearch.github.io/counterfactuals-for-hybrid-systems 14https://github.com/Basis Research/chirho |
| Open Datasets | No | We focus on a hypothetical fishery involving three trophic levels apex predators, intermediate predators (the fished species), and forage fish with dynamics captured by the differential equations presented by Zhou & Smith (2017). This indicates the use of a model from prior work, not a specific, publicly available dataset with concrete access information. |
| Dataset Splits | No | The paper is theoretical and uses a hypothetical model for a case study, rather than empirical data that would require training, testing, or validation splits. No dataset splits are mentioned. |
| Hardware Specification | Yes | The simulated results presented here ran on a consumer grade laptop in the order of one hour. |
| Software Dependencies | No | Code is available here,13 and relies on the dynamical systems package from the causal probabilistic programming language Chi Rho (Basis-Research, 2025).14. While "Chi Rho" is mentioned, no specific version number is provided for this software dependency or any other library/tool. |
| Experiment Setup | No | The paper describes a case study using a hypothetical fishery model and its parameters (e.g., "Let θh2 Np.7, .07q and θh3 Np.07, .007q be elements of θ", "Suppose γ 130 units"). These are model specifications for the case study illustration rather than experimental setup details, hyperparameters, or training configurations for an algorithm or empirical evaluation. |