Litcius/Paper detail

LightRot: A Light-Weighted Rotation Scheme and Architecture for Accurate Low-Bit Large Language Model Inference

Sangjin Kim, Yuseon Choi, Jungjun Oh, Byeongcheol Kim, Hoi‐Jun Yoo

2025IEEE Journal on Emerging and Selected Topics in Circuits and Systems6 citationsDOI

Abstract

As large language models (LLMs) continue to demonstrate exceptional capabilities across various domains, the challenge of achieving energy-efficient and accurate inference becomes increasingly critical. This work presents LightRot, a lightweight rotation scheme and dedicated hardware accelerator designed for low-bit LLM inference. The proposed architecture integrates Grouped Local Rotation (GLR) and Outlier Direction Aligning (ODA) algorithms with a hierarchical Fast Hadamard Transform (FHT)-based rotation unit to address key challenges in low-bit quantization, including the energy overhead of rotation operations. The proposed accelerator, implemented in a 28nm CMOS process, achieves a peak energy efficiency of 27.4 TOPS/W for 4-bit inference, surpassing prior state-of-the-art designs. Unlike conventional approaches that rely on higher-precision inference or evaluate on basic language modeling tasks like GPT-2, LightRot is optimized for advanced models such as LLaMA2-13B and LLaMA3-8B. Its performance is further validated on MT-Bench, demonstrating robust applicability to real-world conversational scenarios and redefining benchmarks for chat-based AI systems. By synergizing algorithmic innovations and hardware efficiency, this work sets a new paradigm for scalable, low-bit LLM inference, paving the way for sustainable AI advancements.

Topics & Concepts

InferenceComputer scienceScheme (mathematics)ArchitectureRotation (mathematics)Bit (key)AlgorithmArithmeticTheoretical computer scienceArtificial intelligenceMathematicsComputer networkMathematical analysisVisual artsArtNatural Language Processing TechniquesTopic ModelingSpeech Recognition and Synthesis