Deeper is Not Always Better: Mitigating the Alignment Tax via Confident Layer Decoding Paper โข 2606.21906 โข Published 18 days ago โข 24 โข 13
Iwaku-Real/Qwen3-0.6B-Base-heretic-test Text Generation โข 0.6B โข Updated 17 days ago โข 35 โข 2
view post Post 10979 THIS IS CRAZY! THE MODEL ON THE IMAGE(Supra-50M-Reasoning) answered correctly and its QUANTIZED IN 2BIT! THE RESPONSE IS CORRECT, IN A 15MB SIZE FILE! See translation 14 replies ยท ๐ฅ 32 32 ๐ 11 11 ๐ง 4 4 ๐ 2 2 + Reply