LLM Connections Solver
Evaluating LLM abstract reasoning through the New York Times Connections word game
About
An evaluation framework and automated solver designed to test the abstract reasoning capabilities of Large Language Models using the NYT Connections game. The repository includes a dataset of 442 games, LLM performance results, and a scoring system based on a knowledge taxonomy.
Details
- Built with
- Unknown
- Creator
- Source date
- Published on X Jun 26, 2024
- Listed
- Added to Dropday just now
- Evidence
- Strong
The page is a verified GitHub repository containing extensive code, datasets, and a link to a formal research paper on arXiv.
Source post
Timeline
Teaser
Video
Playable
Product
Loading…
Media & coverage
sourced from 1 post
