AI 风险行为基准测试：评估 LLM 的风险偏好与一致性

Benchmarking AI risk behaviors

AI 导读

研究团队构建了一套评估 LLM 风险倾向的工具包，可测量模型在不同情境下的风险厌恶或风险容忍程度，以及行为的稳定性与一致性，并支持后续接入新的风险偏好评测。结果显示，2026 年的 SOTA 模型比旧模型更稳定、更一致、更风险中性，原因在于新模型更倾向以期望值最大化作为决策方式。

正文 · 原文

View PDF HTML (experimental)

Abstract:We evaluate LLM risk propensities across a range of contexts and task demands. We evaluate models for how risk averse or risk tolerant they are in various conditions, as well as how stable and consistent their behavior is across conditions. We present a toolkit for developing model risk profiles that can be used to predict model behavior and which can be adapted to include new evaluations of risk propensities. We also examine how these properties have changed across model generations and families. We observe that 2026 state-of-the-art models tend to be more stable, more consistent, and more risk neutral than older models. These trends appear to be related; newer models are more stable and consistent because they are more likely to use expected value maximization as a decision procedure.

Subjects:	Artificial Intelligence (cs.AI); Machine Learning (cs.LG)
Cite as:	arXiv:2611.16318 [cs.AI]
	(or arXiv:2611.16318v1 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2611.16318 arXiv-issued DOI via DataCite (pending registration)

Submission history

From: Luhan Mikaelson [view email]
[v1] Thu, 12 Nov 2026 05:27:44 UTC (1,562 KB)

来源：arXiv cs.AI · arxiv.org