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
- 304
- duration
- 1h 26m
- recorded
- 2026-08-03
run iddclm-nano-curation__4yt6SE3
Score
reward
329.880
Summary
Completed and validated:
- selection.json — 12,967 unique, in-range IDs; exactly 12M packed tokens
- curate.py — byte-identical reproduction confirmed
- claim.md
Best dev perplexity: 325.2, versus 476.7 random baseline. Selection is approximately 3M tokens per target domain.