{
  "hash": "0xcf5d152f515f78c30696a0a54677fb981e7518b129125e958f657f5ddc5de509",
  "content": "<figure>\n  <img src=\"/arc-header.png\" />\n  <figcaption>Mini-ARC: Solving Abstraction and Reasoning Puzzles with Small Transformer Models</figcaption>\n</figure>\n\n# Mini-ARC\n\nOver the last few months, I've spent a lot of time working on strategies for the [ARC Prize](https://arcprize.org/). I'm very motivated to understand and contribute to AI research, and the ARC Prize is the perfect scope to up-skill myself. The [Abstraction and Reasoning Corpus (ARC)](https://arxiv.org/abs/1911.01547) is a benchmark published in 2019 by François Chollet which aims to test artificial systems' ability to reason and efficiently learn new patterns. The benchmark is composed of 2D puzzles that are relatively simple for humans to solve but which have stumped even the most advanced LLMs.\n\nRead more here about my approach: [Mini-ARC: Solving Abstraction and Reasoning Puzzles with Small Transformer Models](https://www.paulfletcherhill.com/mini-arc.pdf).\n",
  "signature": "0x06f5be37dd2dc88ee2ced33e2fc818d88d37c9604cffb1f7b0cb89af4cdf0705adec1b174e1530487a9c60ec05b9f17d8b0ead712cea5cb4f4236c2eb66d8906",
  "date": "2024-11-13T03:23:57.867Z"
}