Skip to content
/ dialop Public

DialOp: Decision-oriented dialogue environments for collaborative language agents

Notifications You must be signed in to change notification settings

jlin816/dialop

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

3 Commits
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

📠 The DialOp Environments

Collaborative decision-oriented dialogue environments for humans and LLM agents.

These environments are associated with the paper Decision-Oriented Dialogue For Human-AI Collaboration.

[Paper] [Website]

Play against GPT-3

Get started and play the games with GPT-3 as an assistant:

pip install -e .
vim dialop/.api_key
python dialop/play_gpt3.py --game {optimization, planning, mediation}

Overview

  • games: define game logic, environment data generation
  • envs: text-based interfaces for games
  • apps: web app interfaces for games for human-human play
Name Description Human UI Agent Text Interface Hosted UI
Optimization Assign the best reviewers for conference papers Optimization UI Optimization Text Interface Play now
Planning Plan the best itinerary with the least travel and price Planning UI Planning Text Interface Play now
Mediation Coordinate a set of group flights that works well for everyone Mediation UI Mediation Text Interface Play now

To play against yourself with the text-based environment:

python dialop/play.py --game {optimization, planning, mediation}

To play against yourself with the web UI:

cd dialop/apps
game={optimization, planning, mediation} flask run

For Planning, you'll need to input a MapboxGL access key in static/client.js and static/map.js to show the map on the agent side and access other map features.

Human-Human Dialogues

To see human-human dialogues:

unzip dialop/data.zip

Self-Play and Prompted Self-Play Evaluation

Coming soon!

Citation

@article{lin2023decision,
  title={Decision-Oriented Dialogue for Human-AI Collaboration},
  author={Lin, Jessy and Tomlin, Nicholas and Andreas, Jacob and Eisner, Jason},
  journal={arXiv preprint arXiv:2305.20076}
  year={2023}
}