IgorGanapolsky's picture
docs: link preprint paper repo
2d989b8 verified
|
Raw
History Blame Contribute Delete
929 Bytes
metadata
pretty_name: ThumbGate Agent Feedback Sample
license: mit
task_categories:
  - text-classification
  - reinforcement-learning
tags:
  - agents
  - feedback
  - dpo
  - thumbgate
  - governance
  - pretooluse
size_categories:
  - n<1K

ThumbGate Agent Feedback Sample

Synthetic / redacted sample of the ThumbGate feedback to DPO export shape.

This is not production customer data. Schema matches:

npx thumbgate export:hf

Files

  • preferences.jsonl — chosen/rejected preference pairs (DPO-ready)
  • traces.jsonl — agent feedback traces (signal + context + outcome)

Product

MIT.

Paper