arXiv cs.AI· Jiashen Ren, Wenlin Zhang, Bohan Zhang, Xiaopeng Li, Zichuan Fu, Wanyu Wang, Junyi Li, Xiangyu Zhao·· 1 天前AI 评分36
LLM 用户模拟中的“中立性缺口”：Persona Following 并非选择性控制
Persona Following Is Not Selective Control: The Neutrality Gap in LLM User Simulation
AI 导读
研究在 8 个黑盒模型和 5 个开放权重模型上发现，仅改变目标属性会连带改变未指定的非目标选择，显式指定非目标方向可几乎消除这种跨属性漂移，但明确声明“无偏好”时仍有 37.7%-76.3% 的留出世界和 51%-81% 的独立非目标项随目标属性漂移，作者称之为“中立性缺口”。
正文 · 原文
Abstract:Persona prompting is widely used in LLM user simulation, yet it relies on an untested assumption: specifying a target user attribute should shift target behavior while leaving other fixed attributes unchanged. Across eight black-box and five open-weight models, however, changing a target attribute alone also shifts unspecified non-target choices, such as risk preference shifting color choice. We test whether explicitly specifying non-target attributes restores selective control across 477 held-out synthetic worlds and 200 independent items under unspecified, directional, and neutral states. Assigning a clear non-target direction nearly eliminates cross-attribute shifts. Explicitly declaring no preference, however, still leaves 37.7%-76.3% of held-out worlds and 51%-81% of independent non-target items shifting with the target attribute while preserving target efficacy. We call this disparity the neutrality gap. Because directional tests can be passed simply by following a stated preference, the neutral state reveals selective-control failures they miss: following a stated persona does not imply keeping unrelated choices invariant. In simulated investment decisions, incidental brochure color labels shift estimated target risk treatment effects by 25.5 percentage points, even when models accurately report having no color preference. Counterfactual attention interventions show that unwanted influence enters primarily at query time, when option evaluation reads the target persona directly. Blocking query-time attention cuts non-target drift by 72.1%-87.8%, whereas directional declarations suppress drift primarily by overwriting option values rather than rerouting attention. Because target and non-target choices share query-time access, blocking it weakens target efficacy while value overwrite imposes an artificial direction, leaving task-dependent selective access as a central challenge.
| Comments: | 61 pages, 9 figures |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2609.35036 [cs.AI] |
| (or arXiv:2609.35036v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2609.35036 arXiv-issued DOI via DataCite |
Submission history
From: Jiashen Ren [view email]
[v1]
Mon, 28 Sep 2026 12:34:21 UTC (921 KB)
[v2]
Thu, 12 Nov 2026 05:17:19 UTC (977 KB)
来源：arXiv cs.AI · arxiv.org