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
- 180
- duration
- 39m
- recorded
- 2026-08-04
run iddclm-nano-curation__frqBRu3
Score
reward
324.110
Summary
Completed and validated:
- selection.json: 16,564 unique valid IDs; packs exactly 12M tokens.
- curate.py: deterministic, label-free quality and domain balancing.
- claim.md: required hypothesis, mechanism, falsification, and transfer.
Measured dev perplexity: 375.833, versus 476.714 for random selection.