← back
x.comÆON FORGE ✨Thu, Jun 4, 2026, 1:47 AM PDT
score 18.0
499likes39RT21reply

Modified Gemma AI model outperforms original on coding tests

Original: Built an Uncensored / Abliterated version of Gemma 4 12B that actually scored HIGHER on OpenAI Human Eval CODING over the base official 12B model.

Source: x.com

Writing ELI5 summary…