@inproceedings{furniturewala-etal-2024-impact,
title = "Impact of Decoding Methods on Human Alignment of Conversational {LLM}s",
author = "Furniturewala, Shaz and
Jaidka, Kokil and
Sharma, Yashvardhan",
editor = "De Clercq, Orph{\'e}e and
Barriere, Valentin and
Barnes, Jeremy and
Klinger, Roman and
Sedoc, Jo{\~a}o and
Tafreshi, Shabnam",
booktitle = "Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, {\&} Social Media Analysis",
month = aug,
year = "2024",
address = "Bangkok, Thailand",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.wassa-1.22",
pages = "273--279",
abstract = "To be included into chatbot systems, Large language models (LLMs) must be aligned with human conversational conventions. However, being trained mainly on web-scraped data gives existing LLMs a voice closer to informational text than actual human speech. In this paper, we examine the effect of decoding methods on the alignment between LLM-generated and human conversations, including Beam Search, Top K Sampling, and Nucleus Sampling. We present new measures of alignment in substance, style, and psychometric orientation, and experiment with two conversation datasets. Our results provide subtle insights: better alignment is attributed to fewer beams in Beam Search and lower values of P in Nucleus Sampling. We also find that task-oriented and open-ended datasets perform differently in terms of alignment, indicating the significance of taking into account the context of the interaction.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="furniturewala-etal-2024-impact">
<titleInfo>
<title>Impact of Decoding Methods on Human Alignment of Conversational LLMs</title>
</titleInfo>
<name type="personal">
<namePart type="given">Shaz</namePart>
<namePart type="family">Furniturewala</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kokil</namePart>
<namePart type="family">Jaidka</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yashvardhan</namePart>
<namePart type="family">Sharma</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-08</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis</title>
</titleInfo>
<name type="personal">
<namePart type="given">Orphée</namePart>
<namePart type="family">De Clercq</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Valentin</namePart>
<namePart type="family">Barriere</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeremy</namePart>
<namePart type="family">Barnes</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Roman</namePart>
<namePart type="family">Klinger</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shabnam</namePart>
<namePart type="family">Tafreshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Bangkok, Thailand</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>To be included into chatbot systems, Large language models (LLMs) must be aligned with human conversational conventions. However, being trained mainly on web-scraped data gives existing LLMs a voice closer to informational text than actual human speech. In this paper, we examine the effect of decoding methods on the alignment between LLM-generated and human conversations, including Beam Search, Top K Sampling, and Nucleus Sampling. We present new measures of alignment in substance, style, and psychometric orientation, and experiment with two conversation datasets. Our results provide subtle insights: better alignment is attributed to fewer beams in Beam Search and lower values of P in Nucleus Sampling. We also find that task-oriented and open-ended datasets perform differently in terms of alignment, indicating the significance of taking into account the context of the interaction.</abstract>
<identifier type="citekey">furniturewala-etal-2024-impact</identifier>
<location>
<url>https://aclanthology.org/2024.wassa-1.22</url>
</location>
<part>
<date>2024-08</date>
<extent unit="page">
<start>273</start>
<end>279</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Impact of Decoding Methods on Human Alignment of Conversational LLMs
%A Furniturewala, Shaz
%A Jaidka, Kokil
%A Sharma, Yashvardhan
%Y De Clercq, Orphée
%Y Barriere, Valentin
%Y Barnes, Jeremy
%Y Klinger, Roman
%Y Sedoc, João
%Y Tafreshi, Shabnam
%S Proceedings of the 14th Workshop on Computational Approaches to Subjectivity, Sentiment, & Social Media Analysis
%D 2024
%8 August
%I Association for Computational Linguistics
%C Bangkok, Thailand
%F furniturewala-etal-2024-impact
%X To be included into chatbot systems, Large language models (LLMs) must be aligned with human conversational conventions. However, being trained mainly on web-scraped data gives existing LLMs a voice closer to informational text than actual human speech. In this paper, we examine the effect of decoding methods on the alignment between LLM-generated and human conversations, including Beam Search, Top K Sampling, and Nucleus Sampling. We present new measures of alignment in substance, style, and psychometric orientation, and experiment with two conversation datasets. Our results provide subtle insights: better alignment is attributed to fewer beams in Beam Search and lower values of P in Nucleus Sampling. We also find that task-oriented and open-ended datasets perform differently in terms of alignment, indicating the significance of taking into account the context of the interaction.
%U https://aclanthology.org/2024.wassa-1.22
%P 273-279
Markdown (Informal)
[Impact of Decoding Methods on Human Alignment of Conversational LLMs](https://aclanthology.org/2024.wassa-1.22) (Furniturewala et al., WASSA-WS 2024)
ACL