# callummcdougall/ARENA_3.0

Repository: https://github.com/callummcdougall/ARENA_3.0
Canonical: https://ross.abutalabs.com/products/arena_30
Language: Jupyter Notebook
License Family: other
Last push: 2026-09-02T13:51:00+00:00

## Health v2 (maintenance only)
Score: 72/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 100, release rhythm 35, longevity 75
- inputs: {"age_days": 1060, "days_push": 0, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases, no_license
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1257, forks 813 (observed 2026-09-03T02:15:09.780183+00:00)

## What it is
ARENA is a hands-on curriculum of exercises and Streamlit pages covering deep learning fundamentals, transformer interpretability, RL, and generative models, aimed at AI alignment researchers. The repo hosts the exercise notebooks and install scripts for the program.

## Use cases
- learn to build a transformer from scratch
- practice mechanistic interpretability with TransformerLens
- implement backpropagation and convolutions in PyTorch
- train GANs and VAEs for image generation
- learn reinforcement learning from scratch
- prepare for AI alignment research
- self-study deep learning fundamentals with exercises

## When to choose
- you want structured, exercise-driven learning in PyTorch and transformer interpretability
- you are preparing for ML or AI safety research
- you prefer hands-on coding over passive tutorials

## When to avoid
- you need production-ready ML software rather than educational material
- you want a maintained library with an API or stable release guarantees
- you need a license permitting redistribution

## Facets
- artifact type: learning-resource
- maturity: active
- function: machine-learning, deep-learning, llm-training, nlp
- domain: machine-learning, deep-learning, large-language-models, tutorials, education
- platform: python, cross-platform
- tags: curriculum, exercises, transformer-interpretability, mechanistic-interpretability, reinforcement-learning, pytorch, streamlit, ai-safety, notebooks

## Member repositories
- callummcdougall/ARENA_3.0 (main) score 72

## Provenance
- Observed fields: from GitHub, fetched 2026-09-03T02:15:09.780183+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T05:07:51.252800+00:00, confidence not recorded.
  - readme: https://github.com/callummcdougall/ARENA_3.0 (fetched 2026-09-03T02:15:09.780183+00:00, sha 330ec0814da7)
- Data as of 2026-08-30T08:39:29.467469+00:00.