Skip to content
Snippets Groups Projects
Commit 4e18a3f5 authored by Maxime Morge's avatar Maxime Morge :construction_worker:
Browse files

PyGAAMAS: Minor corrections

parent 9c17189d
No related branches found
No related tags found
No related merge requests found
......@@ -32,3 +32,6 @@ parameter-efficient fine-tuning, merit further exploration.
% investigation, including test-time scaling methods to enhance reasoning during
% inference, and supervised fine-tuning approaches such as parameter-efficient
% fine-tuning.
Future work should also tackle broader challenges, such as the scalability of
reasoning mechanisms, the transparency of agent behaviors, and the
decentralization of simulations.
File deleted
......@@ -3,7 +3,7 @@
Generative Agents (GAs), powered by Large Language Models (LLMs), have recently
demonstrated remarkable capabilities in simulating human
behaviours~\cite{gao24hssc,guo24arxiv,mei24pnas,wang24fcs}.% reasoning, decision-making
behaviours~\cite{gao24hssc,guo24arxiv,mei24pnas,wang24fcs}. % reasoning, decision-making
These advancements have sparked growing interest in their application to
computational social sciences, where researchers aim to model, analyse, and
predict social dynamics using agent-based
......
......@@ -294,7 +294,7 @@ doi = {10.1177/1043463195007001004}
}
@misc{guo23arxiv,
title={GPT in Game Theory Experiments},
title={{GPT in Game Theory Experiments}},
author={Fulin Guo},
year={2023},
archivePrefix={arXiv},
......
No preview for this file type
......@@ -106,11 +106,10 @@ Matching Pennies game, where opponent strategies are both minimal and credible.
% arbitrary strategies in the Rock-Paper-Scissors game to illustrate this
% behavioral flaw, we adopt the simpler Matching Pennies games, where the opponent
% strategies are both minimal and realistic.
Moreover, their study is not
reproducible: the code is unavailable, and the experiments rely exclusively on
proprietary LLMs. In contrast, we make our code, prompts,
and datasets openly available, and we focus on open-weight
models that can run on standard hardware. %~\cite{pygaamas}
Moreover, their study is not reproducible: the code is unavailable, and the
experiments rely exclusively on proprietary LLMs. In contrast, we openly release
our code, prompts, and datasets, and focus on open-weight models that enable
reproducibility on standard hardware. %~\cite{pygaamas}
% , which required significant computational resources and resulted in
% substantial carbon costs. Furthermore, we reduce environmental impact by
% prompting LLMs to generate algorithmic strategies, as in~\cite{willis25arxiv},
......@@ -130,5 +129,3 @@ coordination, it often creates ambiguity and hinders long-term alignment.
% introduces ambiguity, leading to misaligned expectations and degraded long-term
% coordination.
......@@ -66,7 +66,7 @@ def plot_metric_combined(df_input, metric, ylabel, title, filename, ylim):
plt.grid(True)
plt.legend(title="Model & Messages", loc="best")
plt.tight_layout()
plt.savefig(os.path.join(FIGURE_DIR, f"{filename}.svg"), format="svg")
plt.savefig(os.path.join(FIGURE_DIR, f"{filename}.svg"), format="pdf")
plt.show()
# Plot average payoff (combined)
......
0% Loading or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment