A comprehensive index of artificial intelligence and machine-learning research with AI-generated summaries, citation metrics, and direct links to papers and code.
B. Chen, Gregory W. Wornell
Introduces quantization index modulation (QIM) and distortion-compensated QIM for provably good digital watermarking with optimal rate-distortion-robustness tradeoffs.
Tanishq Kumar, Zachary Ankner, B. Spector, et al.
Proposes precision-aware scaling laws for language models, showing low precision reduces effective parameter count and post-training quantization can harm models trained on more data.
Unknown
QLoRA enables fine-tuning of large language models on a single GPU by combining 4-bit NormalFloat quantization, double quantization, and paged optimizers, matching 16-bit performance.
D. Du, Gu Gong, Xiaowen Chu
A comprehensive survey of model quantization and hardware acceleration techniques for vision transformers.
Ruikang Liu, Haoli Bai, Haokun Lin, et al.
Intactkv improves large language model quantization by preserving pivot tokens, achieving state-of-the-art results.
Jinuk Kim, Marwa El Halabi, Wonpyo Park, et al.
GuidedQuant improves post-training quantization of large language models by using end-loss guidance to better preserve model accuracy.
Unknown
This paper introduces Llmc, a versatile toolkit for benchmarking LLM quantization methods, enabling standardized evaluation and comparison.
Tianyi Zhang, Anshumali Shrivastava
Leanquant introduces a loss-error-aware grid quantization method for LLMs that achieves accurate and scalable compression by minimizing quantization error with respect to the model's loss function.
Yuhang Li, Mingzhu Shen, Yan Ren, et al.
Mqbench provides a reproducible and deployable benchmark for model quantization to accelerate deep learning inference.
Xing Hu, Yuan Cheng, Dawei Yang, et al.
Ostquant improves LLM quantization by applying orthogonal and scaling transformations to better fit weight distributions, reducing accuracy loss.
Unknown
Proposes a retraining-free model quantization method using one-shot weight-coupling learning to achieve efficient compression without fine-tuning.
Kai Liu, Qian Zheng, Kaiwen Tao, et al.
This survey provides a comprehensive overview of low-bit model quantization techniques for deep neural networks, including a curated list of resources.