apol/spain-reference-personas-frontier

Spain Reference Personas Frontier

Spain Reference Personas Frontier is an open synthetic reference population and benchmark substrate for evaluating and designing socially grounded AI systems for Spain.

It is not observed microdata, not a survey, not a prediction of real citizens, and not a substitute for fieldwork, administrative data, or domain-specific validation.

The package is designed for simulation, evaluation, prompt conditioning, subgroup analysis, service design, localization evaluation, and benchmark development for AI systems operating in Spain.

The release is inspired by NVIDIA's Nemotron Personas line, especially the multi-view packaging visible in Nemotron Personas USA, while extending that idea into a benchmark-oriented reference package for Spain.

Package architecture

This release is not a single persona table. It is a layered package:

Responsible use

Use this dataset for controlled simulation, evaluation, service design, research prototyping and benchmark development.

Do not use it for surveillance, immigration enforcement, individual-level persuasion, political microtargeting, voter manipulation, eligibility decisions, credit, employment, housing decisions, or claims about real citizens.

Synthetic personas are not real people and must not be treated as real respondents, real voters, real customers or real service users.

Release snapshot

Why this package exists

• —Controllability: structured fields support filtering, weighting, and subgroup analysis directly.
• —Behavioral usefulness: the release includes actor state and benchmark tasks rather than only descriptive prose.
• —Token efficiency: every public view has a declared budget and measured compliance.
• —Reproducibility: release metadata, task splits, replay seeds, and evaluation summaries are explicit.
• —Household realism: adults remain linked to tenure, burden, caregiving, and consumption context.

Use cases

Example programs

• —Housing policy reaction studies split by tenure, burden, age, and region.
• —Synthetic polling stress-tests that compare short survey answers against richer policy views.
• —Consumer trade-down simulations under inflation using price sensitivity and household constraints.
• —Regional culture and language robustness tests for Spanish-first models serving co-official-language contexts.
• —Event-reaction experiments where the same stable persona receives different recent-media states.
• —Multi-turn family, workplace, or community interaction tasks that require stable persona identity plus mutable memory.

What ships

Package logic

1. 1.Filter or weight cohorts in persona_core.
2. 2.Join household_core when housing or economic context matters.
3. 3.Attach the smallest useful persona view from persona_views.
4. 4.Add actorstateinit only when recency, mood, or event exposure matter.
5. 5.Score behavior with benchmark_tasks instead of relying on anecdotal prompt outputs.

Evaluation at a glance

Observed regional shares

~~~text Andalucía 17.90% ################## Cataluña 16.35% #################- Madrid 14.19% ##############---- Com. Valenciana 10.62% ###########------- Galicia 5.73% ######------------ Castilla y León 5.07% #####------------- País Vasco 4.70% #####------------- Canarias 4.58% #####------------- ~~~

Age calibration summary

View-layer efficiency

~~~text microcard 99.8 / 120 ############-- standardcard 175.7 / 250 ##########---- policyview 89.2 / 180 #######------- consumerview 95.6 / 180 #######------- cultureview 105.8 / 180 ########------ dialogueview 83.4 / 180 ######-------- extended_profile 364.5 / 600 ########------ ~~~

Household and economic context

Benchmark design

Loading

~~~python from datasets import load_dataset

personas = loaddataset( "apol/spain-reference-personas-frontier", "personacore", split="train", token=True, ) ~~~

Limits and cautions

• —This is a synthetic reference population, not observed microdata.
• —The package is suitable for simulation and evaluation, not for replacing field surveys.
• —Age calibration remains the main statistical weakness of v0.1.
• —High-disclosure-tagged rows are exposed as metadata so downstream users can exclude them when needed.
• —Live cross-model benchmark lift is not claimed in the card itself; the bundle ships the infrastructure needed to run it reproducibly.

Companion documents

• —DATASHEET.md
• —METHODS.md
• —SAFE_USE_POLICY.md
• —MODEL_AND_DATA_GOVERNANCE.md
• —EVALUATION_REPORT.md
• —PRIVACY_AND_DISCLOSURE.md
• —EVALUATION_METRICS.json
• —RELEASE_CHECKLIST.md

Citation

Hugging Face DOI: 10.57967/hf/8742

If you use this dataset, please cite:

Guerrero Manchón, M. (2026). Spain Reference Personas Frontier: Synthetic Population and Benchmark Substrate for AI Systems in Spain (v0.1) [Dataset]. Hugging Face / Zenodo. https://doi.org/10.5281/zenodo.20113558

@dataset{guerrero_manchon_2026_spain_reference_personas_frontier,
  author       = {Guerrero Manchón, Miguel},
  title        = {Spain Reference Personas Frontier: Synthetic Population and Benchmark Substrate for AI Systems in Spain},
  year         = {2026},
  publisher    = {Hugging Face and Zenodo},
  version      = {0.1},
  doi          = {10.5281/zenodo.20113558},
  url          = {https://huggingface.co/datasets/apol/spain-reference-personas-frontier}
}