arXiv:2609.16656v1 Announce Type: new Abstract: State space models (SSMs), particularly Mamba, have emerged as efficient alternatives to attention-based architectures and have been extended to vision...
The paper introduces Quantizer‑Aligned Recalibration (QuAR), a single‑pass test‑time adaptation technique for quantized vision transformers that does not require backpropagation or parameter updates. QuAR recalibrates activations at the input of frozen quantizers by aligning per‑channel statistics with the source calibration, thereby correcting the distorted code distribution caused by distribution shift. On ImageNet‑C, QuAR outperforms state‑of‑the‑art backprop‑free methods across 3‑, 4‑, 6‑, and 8‑bit precisions, achieving higher accuracy, lower latency, and minimal memory overhead while maintaining performance across diverse shift scenarios.
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.
arXiv:2610.16368v1 Announce Type: cross Abstract: Looped Language Models (LoopLMs) improve parameter efficiency by recursively reusing shared Transformer modules, yet their repeated computation over...
State space models (SSMs), particularly Mamba, have emerged as efficient alternatives to attention-based architectures and have been extended to vision through ViM, VMamba, and Visual State Space Dual...
MoBiQuant introduces a Mixture-of-Bits quantization framework that allows large language models to be inferred at varying precisions during runtime. It addresses the poor generalization of existing post‑training quantization methods by mitigating precision‑dependent outlier migration through a recursive residual quantization scheme and a token‑aware router that selects the optimal precision per token. Experiments demonstrate that MoBiQuant matches or exceeds single‑precision PTQ performance while providing significant memory savings and up to 1.34× throughput gains over current any‑precision approaches.
arXiv:2509. 10334v2 Announce Type: replace-cross Abstract: Vision Transformers (ViTs) have recently achieved strong results in semantic segmentation, yet their deployment on resource-constrained devices remains limited due to their high memory footprint and computational cost.
SCULPT is a training-time method that enhances the readiness of edge vision models for low-bit post‑training quantization (PTQ). It introduces a topology‑aware activation regularizer to reduce skewness and kurtosis, and a stable percentile‑based clipping mechanism that learns deployment‑ready activation bounds during ordinary FP32 fine‑tuning. The resulting clipping bounds can be directly exported into standard PTQ workflows for INT8 or lower‑bit settings such as W4A8.
arXiv:2608. 11045v1 Announce Type: new Abstract: ReRound (Reconstructive Rounding) is a post-training quantization method that addresses the midpoint ambiguity inherent in standard round-to-nearest (RTN) schemes when quantizing weights near the centers of quantization intervals.
The paper investigates how the choice of reconstruction loss, specifically mean squared error (MSE), affects optimization in sequential post‑training quantization (PTQ) of large language models (LLMs). It identifies an Optimization Imbalance where MSE causes reconstruction loss magnitudes—and thus gradient magnitudes—to vary dramatically across quantization stages, leading to uneven parameter updates. The authors propose that root mean squared error (RMSE) variants, which implicitly normalize gradients, can decouple optimization strength from loss scale and outperform MSE as a drop‑in replacement.
arXiv:2608.21952v1 Announce Type: new Abstract: Recent advances in sequence modeling have highlighted Mamba as a state space architecture offering efficient long-range dependency modeling and providi...
arXiv:2606. 26587v1 Announce Type: cross Abstract: Low-bit floating-point formats and semi-structured sparsity are increasingly supported by modern accelerators, yet combining them for LLM activation compression remains challenging: activations contain input-dependent outliers that dominate block scales in FP4 quantization, and directly applying N:M sparsity masks discards moderate values, coupling sparsification loss with quantization error.