AI Paper Atlas

Research map / Multi-agent systems and games / Human-AI and goal-conditioned

Aligning Machiavellian Agents: Behavior Steering via Test-Time Policy Shaping

Dena Mujtaba, Brian Hu, Anthony Hoogs, Arslan Basharat

AAAI 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Human-AI coordination agents · 3 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add test-time policy shaping with alpha=1.0RL agent · MACHIAVELLI (10 test games)All Violations (normalized): 162.05 -> 94.7±10.1Improved
add test-time policy shaping with alpha=0.5RL agent · MACHIAVELLI (10 test games)All Violations (normalized): 162.05 -> 100.1±4.0Improved
add policy shaping with alpha=1.0RL agent · MACHIAVELLI (10 test games)All Power (normalized): 163.67 -> 87.9±2.0Improved
add policy shaping with alpha=1.0RL agent · MACHIAVELLI (10 test games)Points: 29.67 -> 11.9±1.2Hurt
add policy shaping with alpha=0.5RL agent · MACHIAVELLI (10 test games)Points: 29.67 -> 15.6±0.5Hurt
add policy shaping with alpha=1.0RL agent · MACHIAVELLI (10 test games)Achievements: 14.04 -> 6.5±0.5Hurt
add artificial conscience trainingRL agent · MACHIAVELLI (10 test games)All Violations (normalized): 162.05 -> 105.70Improved
add artificial conscience trainingRL agent · MACHIAVELLI (10 test games)Points: 29.67 -> 27.65Hurt
add ethical prompt to LLM agentLLM agent · MACHIAVELLI (10 test games)All Violations (normalized): 103.58 -> 96.98Improved
add policy shaping to RL-AC agentRL-AC agent · MACHIAVELLIethical violationsMixed
steer RL-AC agent toward violations with higher alphaRL-AC agent · MACHIAVELLIethical violationsImproved
use alpha=1.0 for Killing attributeRL agent · MACHIAVELLI (10 test games)Killing (normalized): 162.21 -> 50.41Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy