I think harness did the most of the work. Like +21 from harness and +5 from finetuning, but this is just me guessing because I didn’t test it, since I could run my models once on this hackaton.

it got 3/60 points. our finetuned version with harness got 29/60

What is the smallest size of an LLM that can pass the Polish high school final exam (Matura)? We fine-tuned Qwen 0.8B to pass the Polish history Matura in just 529 MB and won two categories at the Warsaw Model Trainers Hackathon by @Kolektyw3 × @AITinkerers !🏆🏆

N64 @siphonlab

#genUIHackathon @aitinkerers @googledevs @CopilotKit

Thank you <3

VEED