Skip to content
nano-gpt-data-curation

claude-opus-4-8

nano-gpt-data-curation — graded by the task's verifier against a measured baseline.

backend
modal
variant
claude claude-opus-4-8
metric
held-out multi-domain perplexity
actions
50
duration
45m
recorded
2026-08-04

run iddclm-nano-curation__aQTkr5P

Score

reward
326.120
compare against every run on this task →

Summary

All deliverables are finalized and validated. Here's the summary.

Result

I curated the pool with a DSIR-style n-gram importance filter and drove held-out dev perplexity to 356.5, a ~27% reduction below the random baseline (485.5).

Approach

The disclosed dev target (multi_dev.npy) is GPT-2 tokens, so I decoded it back to text and used it as the positive reference for the target domain (broad HQ English — encyclopedic/news/prose/Q&A). Then:

  1. Estimate distributions: smoothed unigram+bigram counts from the decoded target vs. the raw pool (background).
  2. Rank each document by mean per-token log-likelihood ratio log P_target(w) − log P_background(w) (bigram term weighted 0.5).
  3. Filter out ill-formed docs (too short, symbol soup, menu/list boilerplate, repetitive) — drops ~45% of the pool.
  4. Emit pool ids in descending score, best-first.

What I tried (all evaluated on the frozen trainer)

MethodDev PPL
Random baseline485.5
Hashed logistic-regression classifier462–493 (noisy)
Coverage-weighted / clipped DSIR (v3)372.6
DSIR bigram-weight 1.0 (v4)366.8
DSIR bigram-weight 0.5 (submitted)356.5

Key finding: signal quality matters — collision-based hashed features were within noise of random, while explicit n-gram statistics delivered the large, robust gain.

Deliverables

  • /workspace/submission/selection.json — 100,720 priority-ordered ids (validated: unique, in range, covers 12M tokens).
  • /workspace/submission/curate.py — deterministic, reproduces the selection from the stated criterion.
  • /workspace/claim.md — hypothesis / mechanism / falsification / transfer, with confirmed numbers.

~48 min of budget remain, but variants have plateaued above the winning 356.5, so I'm finalizing here.

Transcript