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..
Evolutionary Prediction Games
Authors: Eden Saig, Nir Rosenfeld
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conclude by complementing our analysis with experiments using both synthetic and real data and coupled with simulated dynamics. Our empirical results shed light on when and how certain user groups are likely to dominate, disappear, or coexist, and demonstrate how different design choices can shape social outcomes even if inadvertently. |
| Researcher Affiliation | Academia | Eden Saig Technion Israel Institute of Technology Haifa, Israel EMAIL Nir Rosenfeld Technion Israel Institute of Technology Haifa, Israel EMAIL |
| Pseudocode | No | The paper does not contain explicit pseudocode or algorithm blocks. Methods are described using mathematical formulations and textual explanations. |
| Open Source Code | Yes | Code is available at: https://github.com/edensaig/evolutionary-prediction-games. |
| Open Datasets | Yes | Method. We use the CIFAR-10 image recognition dataset [36]... Method. We use the MNIST dataset... |
| Dataset Splits | Yes | The training and test sets are split independently, and for each p we measure the prediction model s accuracy on both splits of the test sets (representing acc A(hp) and acc B(hp)). |
| Hardware Specification | Yes | Synthetic data simulations were run on a single Macbook Pro laptop, with 16GB of RAM, M2 processor, and no GPU. Experiments involving neural networks (Section 6) were run on a dedicated server with an AMD EPYC 7502 CPU, 503GB of RAM, and an Nvidia RTX A4000 GPU. |
| Software Dependencies | No | We implement our simulations and analysis in Python. Our synthetic-data experiments rely on scikit-learn [47] for learning algorithm implementations, our CIFAR-10 and MNIST experiments rely on Py Torch [46] and ffcv [38], and our ACSIncome experiment relies on scikit-learn and XGBoost [10]. We use matplotlib [29] for plotting, and mpltern [30] for ternary plots. (No specific version numbers are provided for these software components, only citations to the libraries/papers). |
| Experiment Setup | Yes | Training is performed for 200 epochs using SGD with learning rate 0.01 and momentum 0.5. |