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 [1].

Finding Friend and Foe in Multi-Agent Games

Authors: Jack Serrino, Max Kleiman-Weiner, David C. Parkes, Josh Tenenbaum

NeurIPS 2019 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirical game-theoretic methods show that Deep Role outperforms other hand-crafted and learned agents in five-player Avalon. Deep Role played with and against human players on the web in hybrid human-agent teams. We find that Deep Role outperforms human players as both a cooperator and a competitor.
Researcher Affiliation Collaboration Jack Serrino MIT EMAIL Max Kleiman-Weiner Harvard, MIT, Diffeo EMAIL David C. Parkes Harvard University EMAIL Joshua B. Tenenbaum MIT, CBMM EMAIL
Pseudocode Yes See Appendix A and Alg. 3 for details of the network training algorithm, procedure, parameters and compute details.
Open Source Code Yes Source code and experimental data is available here: https://github.com/Detry322/Deep Role.
Open Datasets No The paper does not provide concrete access information (link, DOI, repository, or formal citation) for a publicly available or open dataset. It describes generating its own training data through self-play and human interactions but does not make this data publicly accessible.
Dataset Splits No The paper does not provide specific dataset split information (percentages, sample counts, or citations to predefined splits) needed to reproduce the data partitioning.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers) needed to replicate the experiment.
Experiment Setup Yes Our network feeds a one-hot encoded vector of the proposer player i and the belief vector b into two fully-connected hidden layers of 80 Re LU units. These feed into a fully-connected win probability layer with sigmoid activation.