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287958d · generated 2026-07-03 · 40 episodes · private draft — not for citation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

Public chat

Experiment 11 — public chat: five principals plus an open channel everyone can read. We OBSERVE what agents use it for — posting rate, and whether they spontaneously report non-cooperators. Written expectation: with a public channel, defection gets called out and promise-violators are punished more than in the private-only game (compare vs experiment 3).

status
planned
coverage
0 / 30 episodes
conditions
pure economy; messages on
config
configs/11_public_chat.yaml

Planned — 30 episodes from 11_public_chat.yaml.

image/svg+xml Matplotlib v3.11.0, https://matplotlib.org/ P L A N N E D 30 episodes · 11_public_chat.yaml

Episodes

No episodes yet — launch with:

uv run python -m delegated_influence.run configs/11_public_chat.yaml