# Sources

The trained `example_2nd_argmax()` checkpoint and task originate in the [Alignment Research Center’s AlgZoo](https://github.com/alignment-research-center/alg-zoo). ARC’s [model post](https://www.alignment.org/blog/algzoo-uninterpreted-models-with-fewer-than-1-500-parameters/) credits Zihao Chen with training and the original analysis.

The checkpoint is distributed under MIT-0. This repository contains its numerical weights in `results/model.json`; the original file SHA-256 is recorded there. It includes no training run or replacement checkpoint.

The circuit account, proofs, figures and demonstrations are by David Khachaturov and released under MIT-0. Derived coefficient tables retain source hashes. The broader validation record distinguishes original and amended outcomes; its large experiment implementation is outside this focused repository.
