nano-gpt-data-curation
gpt-5.6-sol
nano-gpt-data-curation — graded by the task's verifier against a measured baseline.
- backend
- modal
- variant
- codex azure_ai/gpt-5.6-sol
- metric
- held-out multi-domain perplexity
- actions
- 58
- duration
- 19m
- recorded
- 2026-08-03
run iddclm-nano-curation__JnvJaaT
Score
reward
404.830
Summary
Completed all deliverables:
- selection.json: 12,134 unique valid IDs, verified to pack 12M tokens.
- curate.py: reproducible quality scoring and domain balancing.
- claim.md: hypothesis, mechanism, falsification, and transfer.
Submitted selection dev perplexity: 473.4.