LFM2.5-1.2B System One LoRA
Promoted run: seed-only teacher aliases (not grown templates).
- Base:
models/LFM2.5-1.2B-Instruct-8bit(QLoRA on the same 8-bit MLX checkpoint we serve) - Adapter:
adapters/lfm25-1.2b-systemone - Rank 16, scale 2.0, AdamW 5e-5, batch 4, 2 epochs / 64 iters, ~131s
- Targets: attn q/k/v/o + MLP w1/w2/w3, all 16 layers
- Data: 78 train items / 126 examples; frozen holdout 17 items / 29 examples (
evals/splits.json) - Completions match serve-time scoring: option keys, yes/no, level digits. Prompt masked.
Serve (T=1, no calibration.json)
JEVONS_ADAPTER=adapters/lfm25-1.2b-systemone \
JEVONS_TEMPERATURE=1 \
JEVONS_CALIBRATION=off \
JEVONS_MODEL=models/LFM2.5-1.2B-Instruct-8bit \
uv run jevons serve
Or: uv run jevons serve --adapter adapters/lfm25-1.2b-systemone --temperature 1 --calibration off
Disable: --adapter off
T=1 vs previous baseline
| split | run | modal | choice / noul / score | ECE | Brier | acc@≥0.8 | schema |
|---|---|---|---|---|---|---|---|
| full n=155 | base | 69.7% | 78.6 / 68.2 / 56.1 | 0.107 | 0.387 | 92.1% (n=63) | 100% |
| full n=155 | LoRA | 83.9% | 85.7 / 95.5 / 68.3 | 0.077 | 0.193 | 95.9% (n=97) | 100% |
| holdout n=29 | base | 72.4% | 76.9 / 87.5 / 50.0 | 0.122 | 0.390 | 100% (n=11) | 100% |
| holdout n=29 | LoRA | 72.4% | 69.2 / 75.0 / 75.0 | 0.248 | 0.303 | 100% (n=14) | 100% |
Full-set 83.9% includes the 78 train items. Honest ship metric is holdout modal, still 72.4% vs a 90% bar.
Failed grown run (do not serve)
adapters/lfm25-1.2b-systemone-grown: 200 official-Jev labels on grow.py templates + LoRA scale 20. Holdout modal 38%. Choice collapsed to sales / transfer / weather.