← 研踪 PaperTrailarXiv:2602.12316 · v4
GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory
首次提交 2026-02-13 01:29
版本更新 2026-11-11 22:48
雷达观测 2026-11-13 14:16
摘要
Frontier AI systems are increasingly capable and deployed in high-stakes multi-agent environments. However, existing AI safety benchmarks largely evaluate single agents, leaving multi-agent risks such as coordination failure and conflict poorly understood. We introduce GT-HarmBench, a benchmark of 1,535 high-stakes scenarios spanning game-theoretic structures such as the Prisoner's Dilemma, Stag Hunt and Chicken. Scenarios are drawn from realistic AI risk contexts in the MIT AI Risk Repository. Across 15 frontier models, agents fail to choose socially beneficial actions in 39% of high-stakes cases, such as military escalation, election manipulation, and medical malpractice. We measure sensitivity to game-theoretic prompt framing and ordering, and analyze reasoning patterns driving failures. We further show that game-theoretic interventions improve socially beneficial outcomes by up to 31%. Our results highlight substantial reliability gaps and provide a broad standardized testbed for studying alignment in multi-agent environments. The benchmark and code are available at https://github.com/Flecart/gt-harmbench.
作者备注：Accepted at NeurIPS 2026 Main Conference
观测到的版本
相邻记录可能跳过中间版本；首次发现 v2 及以上也会标为版本更新。
v4 · 2026-11-11 22:48 · 最新观测版本
GT-HarmBench: Benchmarking AI Safety Risks Through the Lens of Game Theory
Frontier AI systems are increasingly capable and deployed in high-stakes multi-agent environments. However, existing AI safety benchmarks largely evaluate single agents, leaving multi-agent risks such as coordination failure and conflict poorly understood. We introduce GT-HarmBench, a benchmark of 1,535 high-stakes scenarios spanning game-theoretic structures such as the Prisoner's Dilemma, Stag Hunt and Chicken. Scenarios are drawn from realistic AI risk contexts in the MIT AI Risk Repository. Across 15 frontier models, agents fail to choose socially beneficial actions in 39% of high-stakes cases, such as military escalation, election manipulation, and medical malpractice. We measure sensitivity to game-theoretic prompt framing and ordering, and analyze reasoning patterns driving failures. We further show that game-theoretic interventions improve socially beneficial outcomes by up to 31%. Our results highlight substantial reliability gaps and provide a broad standardized testbed for studying alignment in multi-agent environments. The benchmark and code are available at https://github.com/Flecart/gt-harmbench.
查看此版本解读 →查看此版本原文 ↗