AI News

Wolof‑adapted open‑weight language models (0.6B–14B) — instruction tuning, preference alignment, model merging

ABC2 min read
Model Releases