Memory-light fine-tuning (MeZO) vs AdamW-LoRA, pilot 2
From a pretrained tiny base, AdamW-LoRA reached 0.987-0.995 while MeZO reached 0.55-0.60 of that score (0.9 needed). This pilot separates the methods; MeZO scored almost nothing on the generation task.
Key numbers
MeZO score ratio per seed (need 0.9)
0.60, 0.55, 0.60Source file:
Source hash:
mezo-vs-adam/runs/2026-10-13-pilot2/results.jsonSource hash:
8166e81199f16d5eCaveats
- Tiny 2-layer model, CPU
Watch the roundup
What we tested, and what we learned (90 s)
Video hash: 3de55bc330bba5f8
Request the full write-up
Full write-ups are not public yet. You can request one; requests are reviewed by hand.
Requests are stored privately for up to twelve months, reviewed by hand, never shared, and deleted on request. Contact QMAI@quantummindaillc.com.