FlatQuant: Flatness Matters for LLM Quantization | Read Paper on Bytez