z1_Z_fineweb_edu — Z1 arm Z: 32M trained from scratch on FineWeb-Edu

Why it exists: tests whether training data composition from the first step builds ARC-Easy. Smaller models at the top of the Glint Tiny-ML board were trained from scratch on FineWeb-Edu and score higher on ARC on our evaluation harness than our models trained on ARC-MIX; published small-scale ablations (DataDecide) show the same direction. This arm and arm C (z1_C_arcmix/) use an identical model, recipe and token budget; only the training data differs.

Setup

Model: 31.3M parameters, 15 layers, d_model 384, 6 heads (head dim 64), RoPE, SwiGLU, RMSNorm, QK-norm, value residual; BPE tokenizer with 12,288 tokens (tokenizer.json in the repository root).
Recipe: the 64M flagship recipe scaled down: Muon (hidden 2-D weights) + AdamW, learning rate 6e-4 with 2,000 warmup steps and cosine decay to 6e-5, batch 32 × 1024 tokens, 150,000 steps (4.9B tokens), seed 1337. Training windows are drawn without replacement (each window at most once), less than one epoch.
Data: 5.0B-token pool of FineWeb-Edu (ODC-BY; three sources deduplicated by document hash), scanned against WikiText-2 and ARC test/validation with normalized 13-gram and short-question matching; matching documents were removed.
Checkpoints: every 30,000 steps (about 1B tokens); the last one is step 150,000.
Status: research checkpoints, not a leaderboard submission. The comparison is only between arm Z and arm C; the architecture and trainer differ from our published 32M, so their numbers are not directly comparable.
Format: PyTorch checkpoint dict with model, opt, step, config; train_gpt_ref.py in the repository root rebuilds the model from config.