My 1B model renamed one file to another file's name — the fix was a provenance guard

Description and use

A 1‑billion‑parameter model renamed files to names belonging to other files, causing conflicts. Adding a provenance guard that allows only words from the original filename removed the errors and produced a reliable local file‑tidying tool.

Problem

A 1‑billion‑parameter language model renamed files to names belonging to other files in the same batch, causing cross‑row name collisions and potential data loss.

Outcome

By limiting model‑generated names to those that reuse words from the original filename, the author eliminated cross‑row renaming errors and doubled extensions, achieving a fully reliable file‑tidying workflow with minimal model usage.

Workflow

• Scan folder for loose files
• Apply extension rules
• Apply purpose rules
• Apply meaningful‑name rule
• Batch remaining files and send to local model
• Validate model output: match by position, ensure count, check provenance guard
• Rename files and move to target folders
• Generate plan report
• Apply or undo plan

Evidence

Only the leftovers go to llama3.2:1b: the timestamps, the img.pngs, the AC3.jsons.

Source

The rule I settled on is boring and it works. A proposed name is accepted only if it is built from words that were already in the original filename.

Source

After the guard, the same 599-file run produced zero cross-row names and zero doubled extensions.

Source