Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Parashar, Aditya, Singh, Aditya Vikram, Amballa, Avinash, Lai, Jinlin, Rozonoyer, Benjamin
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2411.06251
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866908338976980992
author Parashar, Aditya
Singh, Aditya Vikram
Amballa, Avinash
Lai, Jinlin
Rozonoyer, Benjamin
author_facet Parashar, Aditya
Singh, Aditya Vikram
Amballa, Avinash
Lai, Jinlin
Rozonoyer, Benjamin
contents Large language models (LLMs) are often equipped with multi-sample decoding strategies. An LLM implicitly defines an arithmetic code book, facilitating efficient and embarrassingly parallelizable \textbf{arithmetic sampling} to produce multiple samples using quasi-random codes. Traditional text generation methods, such as beam search and sampling-based techniques, have notable limitations: they lack parallelizability or diversity of sampled sequences. This study explores the potential of arithmetic sampling, contrasting it with ancestral sampling across two decoding tasks that employ multi-sample inference: chain-of-thought reasoning with self-consistency and machine translation with minimum Bayes risk decoding. Our results demonstrate that arithmetic sampling produces more diverse samples, significantly improving reasoning and translation performance as the sample size increases. We observe a $\mathbf{3\text{-}5\%}$ point increase in accuracy on the GSM8K dataset and a $\mathbf{0.45\text{-}0.89\%}$ point increment in COMET score for WMT19 tasks using arithmetic sampling without any significant computational overhead.
format Preprint
id arxiv_https___arxiv_org_abs_2411_06251
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Quasi-random Multi-Sample Inference for Large Language Models
Parashar, Aditya
Singh, Aditya Vikram
Amballa, Avinash
Lai, Jinlin
Rozonoyer, Benjamin
Artificial Intelligence
Large language models (LLMs) are often equipped with multi-sample decoding strategies. An LLM implicitly defines an arithmetic code book, facilitating efficient and embarrassingly parallelizable \textbf{arithmetic sampling} to produce multiple samples using quasi-random codes. Traditional text generation methods, such as beam search and sampling-based techniques, have notable limitations: they lack parallelizability or diversity of sampled sequences. This study explores the potential of arithmetic sampling, contrasting it with ancestral sampling across two decoding tasks that employ multi-sample inference: chain-of-thought reasoning with self-consistency and machine translation with minimum Bayes risk decoding. Our results demonstrate that arithmetic sampling produces more diverse samples, significantly improving reasoning and translation performance as the sample size increases. We observe a $\mathbf{3\text{-}5\%}$ point increase in accuracy on the GSM8K dataset and a $\mathbf{0.45\text{-}0.89\%}$ point increment in COMET score for WMT19 tasks using arithmetic sampling without any significant computational overhead.
title Quasi-random Multi-Sample Inference for Large Language Models
topic Artificial Intelligence
url https://arxiv.org/abs/2411.06251