Clash of LLMs: A multiagent simulation game to investigate social intelligence of LLMs
This project is based on the concept of red/blue teaming. This concept in prevalent in cyber security and wargaming areas.
We would like to get a simulator where we could configure parameters for our Computer-based agents so that they can play a simulation-based game against each other.
The essential components of this project are:
- a reasonable UI where users can enter parameters
- Ability to connect to various LLMs such as GPT3.5, GPT4, and perhaps Llama (we will cover any cost incured for calling the APIs)
- an underlying simple algorithm that enables the agents to make the choices (please note, this is a software development project, so I am not expecting any high-level algorithmic work, however, I looking for a wrapper that can allow me to deploy a sophisticated model, i..e, by replacing your algorithm ).
- I am also not looking for a 'complex gaming interface'. Rather I am interested in neat display of output which will be the live parameters in the game. This could be in the form of text or graphs.
I have a detailed documentation on what we need. If you are interested then please contact me before picking up this project.
Client
Contact: Mehwish Nasim
Phone: None
Email: [email protected]
Preferred contact: Email
Location: UWA
IP Exploitation Model
The IP exploitation model requested by the Client is: GNU General Public License (open source) http://opensource.org/licenses/GPL-3.0
Department of Computer Science & Software Engineering
The University of Western Australia
Last modified: 16 July 2024
Modified By: Michael Wise
|
 |