arXiv cs.AI· Bingchen Wang, Zi-Yu Khoo, Jingtan Wang·· 14 小时前AI 评分32
P2P：用紧凑 LLM 集成模拟人类偏好，实现人口级调查对齐
Prompts to Proxies: Emulating Human Preferences via a Compact LLM Ensemble
AI 导读
研究者提出 Prompts to Proxies（P2P），一个两阶段推理时系统，通过结构化属性提示构建代理池并用 L1 正则化回归筛选紧凑加权集成，从有限公开调查数据重建目标人群的聚合问卷回答。
正文 · 原文
Abstract:Large language models are increasingly used as proxies for human subjects in social science research, yet external validity requires matching the response distributions of target human populations. We study population-level survey alignment: reconstructing aggregate survey responses from limited public survey data, without individual-level demographic profiles or model finetuning. We formalize this problem as preference reconstruction: rather than matching proxy agents to demographic profiles, we construct a functional basis of proxy agents and recover population preferences by weighted aggregation. We instantiate this idea via Prompts to Proxies ($\texttt{P2P}$), a two-stage inference-time system. Stage 1 uses structured attribute-based prompting with entropy-guided adaptive sampling to construct a diverse proxy pool spanning the latent preference space. Stage 2 employs L1-regularized regression to select a compact weighted ensemble matching observed target-population responses. Across 14 American Trends Panel waves, $\texttt{P2P}$ achieves an average test MSE of 0.014 at approximately 0.8 USD per survey, improving over prompting and demographic-conditioning baselines. On the World Values Survey, cross-locale transfer experiments show that basis expressiveness can outperform locale matching on average, while locale-specific generation helps on culturally divergent questions. A stress test against an SFT-aligned survey model shows competitive performance using less than 3\% of the training data. These results position preference reconstruction as a lightweight, externally verifiable alternative for survey-based population alignment. Code is available at this https URL.
| Comments: | Accepted for publication at NeurIPS 2026 |
| Subjects: | Artificial Intelligence (cs.AI); Computers and Society (cs.CY) |
| Cite as: | arXiv:2509.11311 [cs.AI] |
| (or arXiv:2509.11311v3 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2509.11311 arXiv-issued DOI via DataCite |
Submission history
From: Bingchen Wang [view email]
[v1]
Sun, 14 Sep 2025 15:08:45 UTC (16,061 KB)
[v2]
Wed, 28 Jan 2026 19:39:31 UTC (14,132 KB)
[v3]
Tue, 27 Oct 2026 19:11:00 UTC (14,485 KB)
来源：arXiv cs.AI · arxiv.org