pthinc

/

Cicikus_v4_0.3B_Pitircik

like 1

Follow

PROMETECH BİLGİSAYAR BİLİMLERİ YAZILIM İTHALAT İHRACAT TİCARET ANONİM ŞİRKETİ 11

We fine-tuned the Gemma 0.3B base model using a LoRA-based training approach, achieving an average performance improvement of approximately 50% across our evaluation benchmarks, with a standard deviation of ±5%. This enhancement demonstrates the effectiveness of parameter-efficient fine-tuning in significantly boosting model capability while maintaining low computational overhead.