This work was done during one weekend by research workshop participants and does not represent the work of Apart Research.
ApartSprints
The Concordia Contest: Advancing the Cooperative Intelligence of Language Model Agents
The Concordia Contest: Advancing the Cooperative Intelligence of Language Model Agents
September 9, 2024
Accepted at the 
 research sprint on 
October 3, 2024

Omniscient Narrative Agent

The agents were implemented as predictors of narratives and used this to reflect and analyze their current situation and make decisions to drive the situation for mutual/individual benefit, depending on the environment. Agents were tested across four environment types: reality show scenarios, pub coordination tasks, haggling scenarios and labor collective simulations. Results demonstrated that pre-normalization performance varied significantly across environments, with highest scores in reality show scenarios and lower performance in pub coordination and labor collective tasks. These findings suggest that environmental context substantially influences cooperative behavior, with implications for designing more generalized cooperative AI systems showing Pareto-optimal behavior across all environments.

By 
Jord Nguyen, Akash Kundu, Gayatri K
🏆 
4th place
3rd place
2nd place
1st place
 by peer review
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

This project is private