Under review · project page

Prompts to Proxies

Reconstructing population preferences from a compact LLM ensemble — no finetuning, no demographics, under $1 per survey.

Bingchen Wang · Zi-Yu Khoo · Jingtan Wang

At a glance

From problem to paradigm shift

Problem

Large language models are increasingly used as proxies for human subjects in social science research, but their scientific value depends on external validity: synthetic agents must faithfully reflect the preferences of the populations they are meant to emulate.

Why rethink it

Existing preference-alignment paradigms typically rely on demographic profiling or finetuning a single one-size-fits-all model. These approaches are resource-intensive, often require sensitive data and substantial compute, and do not directly explain how population-level preference heterogeneity should be reconstructed.

Idea

We introduce preference reconstruction theory, which treats preference alignment as a problem of representation learning and statistical inference. Instead of matching profiles, it constructs an expressive basis of proxy agents and recovers population-level preferences through weighted aggregation.

Result

Implemented as a two-stage system, P2P requires no finetuning and no sensitive demographic data, achieves an average test MSE of 0.014 across 14 ATP waves, runs at roughly $0.8 per survey, and yields distinctive empirical insights unavailable under demographic-matching approaches.

How it works

Preference reconstruction in three steps

Animated illustration for Step 1 of P2P

Key results

Efficient, accessible, and empirically validated

0.014 Average test MSE
43% Improvement over best baseline
$0.8 Cost per survey
14 ATP waves evaluated

ATP benchmark

Method Test MSE Entropy Cost
Vanilla 0.026 0.45 $0.7
PERSONA 0.029 0.43 $1.3
P2P 0.014 0.53 $0.8

Interpretation

P2P frames accessibility as a design principle rather than a compromise: no finetuning, no demographic profiling, CPU-friendly deployment, and strong performance under modest resource constraints.

Cost estimates are based on Gemini-2.0-flash.
PERSONA cost only covers the response elicitation stage—higher if generation tokens are included.

Paradigmatic contribution

Beyond digital twins

P2P does not treat faithful emulation as a problem of demographic matching. It treats it as a problem of representation learning and statistical inference.

Preference reconstruction, not profiling.

Existing approaches typically rely on demographic conditioning or a single finetuned model. P2P instead constructs an expressive basis of proxy agents and recovers population-level preferences through weighted aggregation.

Expressive basis beats demographic matching.

Figure 3 shows that a more expressive US-generated basis, after only Stage 2 weight refitting, outperforms HK-native generation on average. This directly challenges the assumption that locale-matched profiling is necessary for faithful emulation.

Localization matters on culturally divergent questions.

The same analysis shows where localization helps: prediction error scales with cultural divergence, and locale-matched generation regains an edge on culturally distinctive dimensions. This turns P2P into a diagnostic tool for studying alignment failure, not just a method for improving fit.

Error analysis comparing transfer and native alignment on Hong Kong WVS.
Figure 3. A more expressive basis generated for the US, after Stage 2 weight refitting, outperforms HK-native generation on average, while localization remains helpful on culturally divergent questions.

Record

Submission history

25 Jul 2025

Submitted to AAAI-26

14 Sep 2025

arXiv v1 posted

18 Sep 2025

Submitted to ICLR-26

28 Jan 2026

arXiv v2 posted

Current

Under review

Citation

BibTeX

@article{wang2025prompts,
            title={Prompts to Proxies: Emulating Human Preferences via a Compact LLM Ensemble},
            author={Wang, Bingchen and Khoo, Zi-Yu and Wang, Jingtan},
            journal={arXiv preprint arXiv:2509.11311},
            year={2025}
            }