Code and Data for the paper
Connecting the Dots: Evaluating Abstract Reasoning Capabilities of LLMs Using the New York Times Connections Word Game
Prisha Samadarshi, Mariam Mustafa, Anushka Kulkarni, Raven Rothkopf, Tuhin Chakrabarty, Smaranda Muresan
Look at
games.txt for all 200 games
knowledge_types.csv for 800 annotations of knowledge types. These are arrange chronologically so each 4 rows = 1 game
automated_call/prompt_llm.txt contains the actual prompt
scoring folder has all the LLM responses
If you use our data or annotations please cite us at
@article{samadarshi2024connecting,
title={Connecting the Dots: Evaluating Abstract Reasoning Capabilities of LLMs Using the New York Times Connections Word Game},
author={Samadarshi, Prisha and Mustafa, Mariam and Kulkarni, Anushka and Rothkopf, Raven and Chakrabarty, Tuhin and Muresan, Smaranda},
journal={arXiv preprint arXiv:2406.11012},
year={2024}
}