· Team project of 4, CS-552 Modern NLP

LLM Teaching Assistant (EPFL-TA-Meister)

Fine-tuned Llama-3-8B with SFT and DPO, raising preference alignment on about 800 held-out EPFL course pairs from 55% to 63%, then shrank it 5.8x with 4-bit GPTQ.

Model → LLM · SFT · DPO · quantization

Released on Hugging Face in full precision and as 8-bit and 4-bit (GPTQ, AWQ, bitsandbytes) models.