Attribution: This repo is a personal study companion to ARENA 3.0 (Alignment Research Engineer Accelerator), a curriculum covering deep learning fundamentals, mechanistic interpretability, RL, evals, and alignment science — originally created by Callum McDougall. This is not official ARENA material. All original curriculum content, exercises, and conceptual design belong to the ARENA authors. Code here is a mix of my own independent implementations and adaptations of the ARENA reference implementations, used for learning purposes. This is a standalone repo rather than a fork because the goal is personal exploration and note-taking, not tracking or contributing back to the original.
- Official curriculum: https://www.arena.education/
- Original repo: https://github.com/callummcdougall/ARENA_3.0
- License of original: No explicit license declared (all rights reserved by the ARENA authors)
uv sync # install dependencies
uv run python <file>Files use # %% cell markers for cell-by-cell execution in VS Code.