hadamard-rotation

Tag

Cards List
#hadamard-rotation

Qift: Shift-Friendly No-Zero W2 Post-Training Quantization for Rotated W2A4/KV4 LLM Inference

arXiv cs.LG · 2026-06-03 Cached

This paper introduces Qift, a fixed no-zero two-bit weight quantization level set designed for Hadamard-rotated LLMs, achieving improved W2A4/KV4 inference by leveraging the near-zero-centered Gaussian-like distribution of rotated weights. Experiments on LLaMA-2-7B and LLaMA-3.1-8B show consistent perplexity gains over standard W2 quantization.

0 favorites 0 likes
#hadamard-rotation

KVarN: Variance-Normalized KV-Cache Quantization Mitigates Error Accumulation in Reasoning Tasks

Hugging Face Daily Papers · 2026-06-02 Cached

KVarN is a calibration-free KV-cache quantizer that uses Hadamard rotation and dual-scaling variance normalization to reduce error accumulation during autoregressive decoding in large language models, achieving state-of-the-art 2-bit precision on reasoning benchmarks.

0 favorites 0 likes
#hadamard-rotation

QAM-W: Joint 2D Codebook Quantization for LLM Weights via Hadamard Rotation and Activation-Aware Scaling

arXiv cs.LG · 2026-05-27 Cached

Introduces QAM-W, a joint 2D codebook quantization method for LLM weights using Hadamard rotation and activation-aware scaling, achieving near BF16 perplexity at 5–6 bits per weight and matching SmoothQuant W8A8 quality with 32% fewer weight bits.

0 favorites 0 likes
← Back to home

Submit Feedback