HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization 事件

HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization arXiv:2605.29843v1 Announce Type: cross Abstract: Post-training quantization (PTQ) is essential for deploying LLMs under memory and bandwidth constraints. However, extreme low-bit quantization remains highly sensitive to activation outliers and anisotropic weight curvature. Existing incoherence-based PTQ methods mitigate this issue with fixed randomized Hadamard transforms (RHTs), which improve quantization r