The paper introduces Fractional Power-of-Two (FPoT) quantization, a base‑√2 weight grid that enables 4‑bit weight quantization for large language models while keeping a fully integer shift‑add datapath. FPoT combines a fractional PoT grid with uniform integer activations, a calibration‑time grid‑alignment refinement, and a lightweight 1‑adder √2 approximation to reduce gate count compared to a 4‑bit multiplier. On Llama‑3‑8B, FPoT boosts zero‑shot commonsense reasoning by 8.72% over the standard PoT baseline and stays within 0.4% of uniform integer performance, with benefits extending across seven models and three bit‑widths.
arXiv:2605. 08692v2 Announce Type: replace Abstract: Post-training weight-only quantization to 4 bits is widely used to reduce the memory and compute costs of large language model inference.
arXiv:2602. 01027v2 Announce Type: replace Abstract: Mixed-precision quantization is a promising approach for compressing large language models under tight memory budgets.
arXiv:2608. 07019v1 Announce Type: new Abstract: Post-training quantization (PTQ) is widely used to reduce the memory and computational cost of large language models.
arXiv:2607. 14618v1 Announce Type: new Abstract: CPUs are the most universal target for on-device LLM inference, but existing low-bit quantization methods offer either coarse operating points or fine-grained mixed precision that is difficult to execute efficiently on CPUs.
Squeeze10-LLM is a staged mixed‑precision post‑training quantization framework that reduces 16‑bit LLM weights to an average of 1.6 bits per weight by assigning 80% of weights to 1 bit and 20% to 4 bits. It introduces Post‑Binarization Activation Robustness (PBAR), a weight significance metric that considers activation impact, and Full Information Activation Supervision (FIAS), a strategy that preserves activation information to limit error propagation. Experiments on LLaMA and LLaMA2 demonstrate that Squeeze10‑LLM achieves state‑of‑the‑art performance for sub‑2‑bit weight‑only quantization, raising average accuracy from 43% to 56% on six zero‑shot classification tasks.
arXiv:2606. 04115v1 Announce Type: cross Abstract: Quantizing large language models (LLMs) to low-precision floating-point representations is central to efficient deployment, yet applying a single bit-width uniformly across all layers is sub-optimal in terms of both performance and accuracy.
arXiv:2606. 13054v1 Announce Type: cross Abstract: Large language models (LLMs) exhibit exceptional general language processing capabilities, but their memory and compute costs hinder deployment.
The paper introduces algebraic integer number formats (AINF) as a framework for highly configurable and accurate compressed computation, emphasizing its reduction to purely integer arithmetic for error‑free multiply‑accumulate operations. A matrix multiplication engine leveraging AINF achieves performance close to the most efficient numerical counterparts on modern hardware accelerators. The authors further propose algebraic integer quantization (AIQ) as a lossy compression format for large language model inference, demonstrating competitive accuracy against the industry‑standard MX format at equivalent bit‑per‑parameter budgets, and suggest that higher‑order algebraic integers could yield even greater computational benefits.
arXiv:2606. 15652v1 Announce Type: new Abstract: 4-bit quantization significantly reduces the memory footprint and accelerates the inference of large language models (LLMs).
arXiv:2606. 00079v1 Announce Type: cross Abstract: Mixture-of-Experts (MoE) large language models reduce per-token computation through sparse expert activation, but their deployment remains memory-intensive because all expert weights must be kept resident in memory.
arXiv:2410. 13056v4 Announce Type: replace-cross Abstract: Large Language Models (LLMs) have demonstrated remarkable success across a wide range of language tasks, but their deployment on edge devices remains challenging due to the substantial memory requirements imposed by their large parameter sizes.
DualQuant introduces a dual‑scale, calibration‑free post‑training quantization framework that jointly optimizes weight‑reconstruction MSE and derives per‑channel activation rescalings, enabling effective 4‑bit weight and activation quantization without calibration data. The method employs a non‑differentiable optimization solved via variable splitting and block coordinate descent, supporting a wide range of formats including MX and NVFP4. Experiments on Llama and Qwen models show DualQuant matches or surpasses calibration‑based SmoothQuant and significantly outperforms other calibration‑free baselines, achieving up to 2.61× memory reduction and 2.46× throughput speedup on an RTX 4090.