Strategic Distribution Shift of Interacting Agents via Coupled Gradient Flows
Authors: Lauren Conger, Franca Hoffmann, Eric Mazumdar, Lillian Ratliff
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Empirically, we show that our approach captures well-documented forms of distribution shifts like polarization and disparate impacts that simpler models cannot capture. |
| Researcher Affiliation | Academia | Lauren Conger California Institute of Technology lconger@caltech.edu Franca Hoffmann California Institute of Technology franca.hoffmann@caltech.edu Eric Mazumdar California Institute of Technology mazumdar@caltech.edu Lillian Ratliff University of Washington ratliffl@uw.edu |
| Pseudocode | No | The paper does not contain structured pseudocode or algorithm blocks. |
| Open Source Code | No | The paper does not provide an explicit statement about the release of its source code or a link to a code repository. |
| Open Datasets | No | The paper uses mathematical models and simulated distributions (e.g., 'Gaussian initial condition') rather than publicly available datasets for its experiments, and thus does not provide access information for a dataset. |
| Dataset Splits | No | The paper conducts numerical simulations based on theoretical models rather than experiments on traditional datasets, and therefore does not specify training, validation, or test splits. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the simulations. |
| Software Dependencies | No | The paper mentions that 'The PDEs were implemented based on the finite volume method from [CCH15]' but does not provide specific software names with version numbers. |
| Experiment Setup | Yes | In Figure 1, we simulate two extremes of the timescale setting; first when ρ is nearly best-responding and then when x is best-responding. The simulations have the same initial conditions and end with the same distribution shape; however, the behavior of the strategic population differs in the intermediate stages. ... The coefficient weights are α = 0.1 and β = 0.05, with discretization parameters dz = 0.1, dt = 0.01. |