Template
Training Data Sheet
Summarize already-cleared training inputs for a model run. This template is not a submission or transfer form.
Training data sheet sections
- Model run or checkpoint ID
- High-level source categories already cleared before training
- License and usage restrictions summary
- Privacy and contamination checks
- Languages, domains, and Taiwan-local coverage
- Mixture weights or sampling notes
- Filtering and preprocessing steps
- Evaluation gates tied to the run
- Known limitations
- Release decision and contact path