LLM Compression: How Far Can We Go in Balancing Size and Performance?

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216208%3A11320%2F25%3A10511654" target="_blank" >RIV/00216208:11320/25:10511654 - isvavai.cz</a>

Výsledek na webu

<a href="https://acl-bg.org/proceedings/2025/RANLP%202025/pdf/2025.ranlp-1.136.pdf" target="_blank" >https://acl-bg.org/proceedings/2025/RANLP%202025/pdf/2025.ranlp-1.136.pdf</a>

DOI - Digital Object Identifier

—

Jazyk výsledku

angličtina

Název v původním jazyce

LLM Compression: How Far Can We Go in Balancing Size and Performance?

Popis výsledku v původním jazyce

Quantization is an essential and popular technique for improving the accessibility of large language models (LLMs) by reducing memory usage and computational costs while maintaining performance. In this study, we apply 4-bit Group Scaling Quantization (GSQ) and Generative Pretrained Transformer Quantization (GPTQ) to LLaMA 1B, Qwen 0.5B, and PHI 1.5B, evaluating their impact across multiple NLP tasks. We benchmark these models on MS MARCO (Information Retrieval), BoolQ (Boolean Question Answering), and GSM8K (Mathematical Reasoning) datasets, assessing both accuracy and efficiency across various tasks. The study measures the trade-offs between model compression and task performance, analyzing key evaluation metrics namely: accuracy, inference latency, and throughput (total output tokens generated per second), providing insights into the suitability of low-bit quantization for real-world deployment. Using the results, a user can then make a suitable decision based on the specifications that need to be

Název v anglickém jazyce

LLM Compression: How Far Can We Go in Balancing Size and Performance?

Popis výsledku anglicky

Quantization is an essential and popular technique for improving the accessibility of large language models (LLMs) by reducing memory usage and computational costs while maintaining performance. In this study, we apply 4-bit Group Scaling Quantization (GSQ) and Generative Pretrained Transformer Quantization (GPTQ) to LLaMA 1B, Qwen 0.5B, and PHI 1.5B, evaluating their impact across multiple NLP tasks. We benchmark these models on MS MARCO (Information Retrieval), BoolQ (Boolean Question Answering), and GSM8K (Mathematical Reasoning) datasets, assessing both accuracy and efficiency across various tasks. The study measures the trade-offs between model compression and task performance, analyzing key evaluation metrics namely: accuracy, inference latency, and throughput (total output tokens generated per second), providing insights into the suitability of low-bit quantization for real-world deployment. Using the results, a user can then make a suitable decision based on the specifications that need to be

Druh

O - Ostatní výsledky

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

<a href="/cs/project/EH23_020%2F0008518" target="_blank" >EH23_020/0008518: Jazykověda, umělá inteligence a jazykové a řečové technologie: od výzkumu k aplikacím</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>I - Institucionalni podpora na dlouhodoby koncepcni rozvoj vyzkumne organizace

Rok uplatnění

2025

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů