A 135M overfit gate, then a token-matched 1B comparison. The decoder uses the concept channel at both scales; the token verdict at 1B is neutral.
🤝 Open to Collab
Varfolomii
yava-code
AI & ML interests
None yet
Recent Activity
liked a dataset 14 days ago
aidigestorg/ai-village updated a model about 1 month ago
yava-code/Tessera-1B-Nano-Base liked a model about 1 month ago
yava-code/Tessera-1B-Nano-BaseOrganizations
