Published

Multilingual Code-Switch Recovery Protocol

Joe Wu
Joe Wu
0

Audited multilingual character catalog, benchmark protocols, and developer-ready feeds for Japanese, Korean, Spanish, Portuguese, and Chinese AI companion research.

Multilingual Code-Switch Recovery Protocol

This open dataset is a blank, reproducible protocol for evaluating persona recovery after language switching. It contains no user data and no claimed benchmark result. The CSV includes twelve ordered test steps, evidence fields, and a blank 0-4 score column.

Files

  • protocol.csv: evaluation steps and empty evidence fields
  • metadata.json: machine-readable Dataset metadata and distribution URLs

Method

  1. Use a fresh test account and a documented model/version.
  2. Follow the steps in order without inserting personal or identifying data.
  3. Record observable behavior and redact excerpts before publication.
  4. Publish failures and incomplete runs as well as successful runs.
  5. Cite this protocol version and disclose any relationship with Ponys.ai.

Product path under test

https://ponys.ai/ai-character-generator

License

CC BY 4.0. Maintained by the Ponys.ai team; not an independent review.