ProfileFoundry: A Synthetic Person-Object Substrate for Privacy, Memory, and Tool-Use Evaluation in LLM Agents
Sriram Selvam Anneswa Ghosh [email protected] [email protected] · 2026
Read ProfileFoundry: A Synthetic Person-Object Substrate for Privacy, Memory, and Tool-Use Evaluation in LLM Agents by Sriram Selvam Anneswa Ghosh [email protected] [email protected] (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2606-26403.txt.
Description
Sriram Selvam and Anneswa Ghosh release ProfileFoundry, a generator of fully synthetic people for testing AI systems that handle personal data. Real user records are hard to share or audit safely, while randomly faked fields lack the consistency serious evaluation needs.
The fixed release holds 100,000 synthetic adult person objects across eight locales, each with a current snapshot, household and employer links, and events aligned over time. In total it carries about 709,000 events and 519,000 relationship links. The authors check population margins, per-object consistency, and provenance, while stating plainly that it is a test substrate, not a model of a real population.
Catalog record
| CREATOR | Sriram Selvam Anneswa Ghosh [email protected] [email protected] |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 61 KB |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2606.26403