AI Paper Atlas

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

The Oversight Game: Learning to Cooperatively Balance an AI Agent's Safety and Autonomy

William Overman, Mohsen Bayati

ICML 2026 · Multi-agent systems and games › Human-AI and goal-conditioned · approach: Human-AI coordination agents

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
train AI and human by independent policy gradient on shared safety rewardwrapped unsafe base policy · Lavalandviolation rate: ~15% -> zeroImproved
train both players over time on shared rewardOversight Game players · Lavalandintervention rate: 50% -> 28.2%Improved
relax team-game to unshared asymmetric costsOversight Game · Lavalandviolation rateNo clear effect
fine-tune LLM policies with independent policy gradientQwen3-30B-A3B LoRA agents · ToolEmu Oversight MDPstest reward: -20 -> +8.4Improved
learned joint policy at risky vs safe statesLLM oversight policies · ToolEmu Oversight MDPsAI ask rate: 99.4% at risky, 3.9% at safeImproved

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