Python command line application for generating text based puzzles for training and testing machine learning models.
Project description
Text based Puzzle Generators for AI Research
This repository contains functions for generating puzzles composed of blocks of English text or ASCII based diagrams.
There is a command line application for interacting with the puzzle generators as well as a library that can be used to incorporate these puzzles in your application of model testing framework.
The puzzles are generated randomly from predetermined distributions. Each of the generators is designed to use text to create higher level representations of a puzzle domain that requires the development of spontaneous mental models in order to manipulate cognitively to reason about and solve the problems.
Some CLI parameters allow for generating from restricted complexty sets. These are used when experimenting with complexity thresholds that affect both human and machine performance on these tasks.
Usage
Citations
This package is being presented at the ICEAST conference in May 2025
@article{hawkins:2025,
author = {John Hawkins},
title = {Enigme: Generative Text Puzzles for Evaluating Reasoning in Language Models}
booktitle = {2025 11th International Conference on Engineering, Applied Sciences, and Technology (ICEAST)}
year = {2025},
month = {05},
}
Additional work on analysis to follow, watch this space.
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
File details
Details for the file enigme-0.0.0.tar.gz.
File metadata
- Download URL: enigme-0.0.0.tar.gz
- Upload date:
- Size: 65.8 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: twine/6.1.0 CPython/3.10.8
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
f7a5f56439c188d39da6fed403fd67b77f25df0098c8b5606743d6ea59cef5f8
|
|
| MD5 |
2bb39946e05dbe3ca87054e0ed5b2157
|
|
| BLAKE2b-256 |
bf6e8135becaa7642eda59e5a730ffd257fb9ee37460a87c828e65ffc344f4e7
|