arXiv cs.AI· Hanming Yang, Daksh Mittal, Jing Dong, Hongseok Namkoong·· 1 天前AI 评分38
通过迭代去对齐估计稀有事件概率
Rare Event Estimation via Iterative Unalignment
AI 导读
研究者提出一种新的重要性采样（IS）方法，通过扰动原模型权重构建可微参数化的语言模型提案，实现权重空间上的梯度搜索，用于估计智能体随机输出轨迹中的稀有事件概率。在约 120M 和约 2.6B 模型、覆盖 300+ 个低至 10⁻⁹ 的稀有事件上评估，对概率低于 10⁻⁷ 的事件，其 IS 估计器平均比朴素蒙特卡洛获得超过 800 倍的计算加权效率提升。
正文 · 原文
Abstract:As agents are deployed with increased autonomy, even extremely rare events along their stochastic output trajectories can occur and prove catastrophic. Safe deployment therefore does not depend on whether these events can occur, but on how often they might. We study the problem of estimating the probability of rare events that arise from stochastic variation in the agent's own actions. Estimating this type of risk requires searching over the combinatorially vast space of trajectories. Naive Monte Carlo is computationally prohibitive in this regime, and constructing effective importance sampling (IS) proposals requires coordinated changes to a context-dependent chain of conditional distributions. We develop a new IS method that perturbs the original model's weights to construct the proposal. The proposal is itself a differentiably parameterized language model, enabling gradient-based search over weight space. We formulate an objective that combines a differentiable surrogate for event amplification and an adaptive regularization scheme that dynamically balances amplification against estimator stability. We evaluate our approach on $\sim$120M and $\sim$2.6B models across three event families spanning 300+ rare events as rare as $10^{-9}$, all with independently estimated reference probabilities. In our most verifiable settings, we observe that our IS estimator achieves over $800\times$ compute-weighted efficiency gains on average over naive Monte Carlo for events with probabilities lower than $10^{-7}$. Our implementation is available at this https URL.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.24969 [cs.LG] |
| (or arXiv:2609.24969v2 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2609.24969 arXiv-issued DOI via DataCite |
Submission history
From: Hanming Yang [view email]
[v1]
Mon, 21 Sep 2026 17:53:38 UTC (2,918 KB)
[v2]
Thu, 12 Nov 2026 04:59:13 UTC (2,905 KB)
来源：arXiv cs.AI · arxiv.org