Skip to content

README.md

Latest commit

 

History

History
61 lines (47 loc) · 1.76 KB

README.md

File metadata and controls

61 lines (47 loc) · 1.76 KB
Raw

Evaluarium — website

Interactive explorer for AI Evaluation Ecosystem simulation runs. Once deployed, will live at https://aimslab.stanford.edu/evaluarium/.

Pages

File What it does
client/index.html Condition card grid (bucket / mode / model / search filters; compare mode)
client/viewer.html Full single-run playback — fetched via ?run=<path>
client/compare.html Multi-run comparison (condition × seed × p10/p90)

Running locally

cd website/client
python -m http.server 8080
# open http://localhost:8080

Pure static — any HTTP server works.

URL formats

index.html?bucket=core_privacy&mode=llm&model=claude-sonnet-4-6&q=baseline
viewer.html?run=core_privacy/llm/claude-sonnet-4-6/baseline/seed_42
compare.html?conditions=baseline|public_only&seeds=seed_42&model=claude-sonnet-4-6
compare.html?conditions=baseline&seeds=seed_42|seed_43&model=claude-sonnet-4-6&agg=1

Data

client/data/ (~65 MB, 367 frame files) and client/runs.json (~500 KB) are bundled in this repo. Provenance lives in client/data/VERSION. Regenerated upstream via scripts/animation/generate_explorer.py.

Layout

website/
├── README.md        ← this file
├── DESIGN.md        ← architecture and tier roadmap
├── DEPLOYMENT.md    ← hosting notes
└── client/          ← pure-static HTML/JS/CSS + bundled run data
    ├── index.html
    ├── viewer.html
    ├── compare.html
    ├── runs.json
    ├── LICENSE
    ├── js/
    │   └── loader.js
    └── data/
        ├── VERSION
        └── <bucket>/<mode>/<model?>/<condition>/seed_<N>/frames.json

License

MIT — see client/LICENSE.