A 27M parameter model beating LLMs on reasoning tasks6 points·by SteadySurfdom·7 miesięcy temu·1 comments