Experiment
Five principals
Experiment 3 — five principals: the 5 models we care about, all playing each other. Small and cheap -> run many times for tight error bars (Q1 persuasion, Q3 leaderboard, Q6 strategies). Expectation: at 30 reps, capture differences between the five principals exceed their error bars.
- status
- planned
- coverage
- 0 / 30 episodes
- conditions
- pure economy; messages on
- config
- configs/03_five_principals.yaml
Planned — 30 episodes from 03_five_principals.yaml.
Episodes
No episodes yet — launch with:
uv run python -m delegated_influence.run configs/03_five_principals.yaml