Skip to content
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
217
duration
52m
recorded
2026-08-04

run iddclm-nano-curation__nfvFpyv

Score

reward
329.220
compare against every run on this task →

Summary

Completed and validated:

  • selection.json: 182,016 unique valid IDs; 12M cutoff reached after 12,443 documents.
  • curate.py: reproducible quality + Moore–Lewis ranking.
  • claim.md: required hypothesis and analysis.

Frozen-trainer dev perplexity: 323.395, versus 476.714 for random selection.

Transcript