Delegated Influence

A competitive multi-agent benchmark for LLM persuasion: the only way to score is to get other agents to spend their scarce actions on you.

287958d · generated 2026-07-03 · 40 episodes · private draft — not for citation

Experiment

Mixed economy

Experiment 7 โ€” mixed economy: pulling your own lever pays 0.5. Q6 act-vs-influence: do models still bother persuading when a safe self-serve option exists? Expectation: weaker models take the safe self-pull more; stronger models keep trading.

status
planned
coverage
0 / 405 episodes
conditions
complete topology; mixed economy (self-pull 0.5); messages on
questions
q6
config
configs/07_attack_mixed.yaml

Planned โ€” 405 episodes from 07_attack_mixed.yaml.

image/svg+xml Matplotlib v3.11.0, https://matplotlib.org/ P L A N N E D 405 episodes ยท 07_attack_mixed.yaml

Episodes

No episodes yet — launch with:

uv run python -m delegated_influence.run configs/07_attack_mixed.yaml