何时单步隐变量分数可替代前瞻搜索？主动信息获取中的深度与描述之辨（arXiv 论文）
Depth or Description? When a One-Step Latent Score Replaces Lookahead in Proactive Information Acquisition
AI 导读
一篇 arXiv 论文研究主动智能体在行动前如何决定提问：精确前瞻搜索开销大，而单步分数虽便宜但可能遗漏答案间的交互。研究发现，单步分数在标签上会“失明”，但在合适的隐变量上评分即可用极小搜索代价替代大部分前瞻深度；同一问题的不同描述方式会改变结论。此外，从一个语言模型答案中学到的概率不能迁移到另一个模型，从 token 读出的概率只值几个标签。
正文 · 原文
Abstract:A proactive agent must decide what to ask before it acts. Exact lookahead values combinations of questions but is expensive and leans on a fitted model; a one-step score is cheap but can miss them. Our main finding: the need for depth often lies in how the hidden cause is described. A one-step score on the label can be blind to an interaction among answers. A score on a suitable latent cause is not, and where that latent encodes the decision it replaces most of depth at a fraction of the search. Described differently, the same problem can defeat the score while exact values do not move, so keep the latent small and check it against exact values. Probabilities learned from one language model's answers do not carry to another unless planning allows for both; those read from tokens are worth only a few labels.
| Subjects: | Machine Learning (cs.LG); Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2611.07155 [cs.LG] |
| (or arXiv:2611.07155v1 [cs.LG] for this version) | |
| https://doi.org/10.48550/arXiv.2611.07155 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Biswa Sengupta [view email]
[v1]
Wed, 4 Nov 2026 21:27:35 UTC (327 KB)
来源：arXiv cs.AI · arxiv.org