@inproceedings{joshi-etal-2025-coke,
title = "{C}o{K}e: Customizable Fine-Grained Story Evaluation via Chain-of-Keyword Rationalization",
author = "Joshi, Brihi and
Venkatapathy, Sriram and
Bansal, Mohit and
Peng, Nanyun and
Chang, Haw-Shiuan",
editor = "Arviv, Ofir and
Clinciu, Miruna and
Dhole, Kaustubh and
Dror, Rotem and
Gehrmann, Sebastian and
Habba, Eliya and
Itzhak, Itay and
Mille, Simon and
Perlitz, Yotam and
Santus, Enrico and
Sedoc, Jo{\~a}o and
Shmueli Scheuer, Michal and
Stanovsky, Gabriel and
Tafjord, Oyvind",
booktitle = "Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM{\texttwosuperior})",
month = jul,
year = "2025",
address = "Vienna, Austria and virtual meeting",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.gem-1.31/",
pages = "366--384",
ISBN = "979-8-89176-261-9",
abstract = "Evaluating creative text such as human-written stories using language models has always been a challenging task {--} owing to the subjectivity of multi-annotator ratings. To mimic the thinking process of humans, chain of thought (Wei et al., 2023) (CoT) generates free-text explanations that help guide a model{'}s predictions and Self-Consistency (Wang et al., 2022) (SC) marginalizes predictions over multiple generated explanations. In this study, we discover that the widely-used self-consistency reasoning methods cause suboptimal results due to an objective mismatch between generating `fluent-looking' explanations vs. actually leading to a good rating prediction for an aspect of a story. To overcome this challenge, we propose Chain-of-Keywords (CoKe), which generates a sequence of keywords before generating a free-text rationale, that guide the rating prediction of our evaluation language model. Then, we generate a diverse set of such keywords, and aggregate the scores corresponding to these generations. On the StoryER dataset, CoKe based on our small fine-tuned evaluation models not only reach human-level performance and significantly outperform GPT-4 with a 2x boost in correlation with human annotators, but also requires drastically less {\#} of parameters."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="joshi-etal-2025-coke">
<titleInfo>
<title>CoKe: Customizable Fine-Grained Story Evaluation via Chain-of-Keyword Rationalization</title>
</titleInfo>
<name type="personal">
<namePart type="given">Brihi</namePart>
<namePart type="family">Joshi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sriram</namePart>
<namePart type="family">Venkatapathy</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Nanyun</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haw-Shiuan</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM²)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Ofir</namePart>
<namePart type="family">Arviv</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Miruna</namePart>
<namePart type="family">Clinciu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaustubh</namePart>
<namePart type="family">Dhole</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rotem</namePart>
<namePart type="family">Dror</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastian</namePart>
<namePart type="family">Gehrmann</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Eliya</namePart>
<namePart type="family">Habba</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Itay</namePart>
<namePart type="family">Itzhak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Simon</namePart>
<namePart type="family">Mille</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yotam</namePart>
<namePart type="family">Perlitz</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enrico</namePart>
<namePart type="family">Santus</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">Michal</namePart>
<namePart type="family">Shmueli Scheuer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gabriel</namePart>
<namePart type="family">Stanovsky</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oyvind</namePart>
<namePart type="family">Tafjord</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Vienna, Austria and virtual meeting</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-261-9</identifier>
</relatedItem>
<abstract>Evaluating creative text such as human-written stories using language models has always been a challenging task – owing to the subjectivity of multi-annotator ratings. To mimic the thinking process of humans, chain of thought (Wei et al., 2023) (CoT) generates free-text explanations that help guide a model’s predictions and Self-Consistency (Wang et al., 2022) (SC) marginalizes predictions over multiple generated explanations. In this study, we discover that the widely-used self-consistency reasoning methods cause suboptimal results due to an objective mismatch between generating ‘fluent-looking’ explanations vs. actually leading to a good rating prediction for an aspect of a story. To overcome this challenge, we propose Chain-of-Keywords (CoKe), which generates a sequence of keywords before generating a free-text rationale, that guide the rating prediction of our evaluation language model. Then, we generate a diverse set of such keywords, and aggregate the scores corresponding to these generations. On the StoryER dataset, CoKe based on our small fine-tuned evaluation models not only reach human-level performance and significantly outperform GPT-4 with a 2x boost in correlation with human annotators, but also requires drastically less # of parameters.</abstract>
<identifier type="citekey">joshi-etal-2025-coke</identifier>
<location>
<url>https://aclanthology.org/2025.gem-1.31/</url>
</location>
<part>
<date>2025-07</date>
<extent unit="page">
<start>366</start>
<end>384</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CoKe: Customizable Fine-Grained Story Evaluation via Chain-of-Keyword Rationalization
%A Joshi, Brihi
%A Venkatapathy, Sriram
%A Bansal, Mohit
%A Peng, Nanyun
%A Chang, Haw-Shiuan
%Y Arviv, Ofir
%Y Clinciu, Miruna
%Y Dhole, Kaustubh
%Y Dror, Rotem
%Y Gehrmann, Sebastian
%Y Habba, Eliya
%Y Itzhak, Itay
%Y Mille, Simon
%Y Perlitz, Yotam
%Y Santus, Enrico
%Y Sedoc, João
%Y Shmueli Scheuer, Michal
%Y Stanovsky, Gabriel
%Y Tafjord, Oyvind
%S Proceedings of the Fourth Workshop on Generation, Evaluation and Metrics (GEM²)
%D 2025
%8 July
%I Association for Computational Linguistics
%C Vienna, Austria and virtual meeting
%@ 979-8-89176-261-9
%F joshi-etal-2025-coke
%X Evaluating creative text such as human-written stories using language models has always been a challenging task – owing to the subjectivity of multi-annotator ratings. To mimic the thinking process of humans, chain of thought (Wei et al., 2023) (CoT) generates free-text explanations that help guide a model’s predictions and Self-Consistency (Wang et al., 2022) (SC) marginalizes predictions over multiple generated explanations. In this study, we discover that the widely-used self-consistency reasoning methods cause suboptimal results due to an objective mismatch between generating ‘fluent-looking’ explanations vs. actually leading to a good rating prediction for an aspect of a story. To overcome this challenge, we propose Chain-of-Keywords (CoKe), which generates a sequence of keywords before generating a free-text rationale, that guide the rating prediction of our evaluation language model. Then, we generate a diverse set of such keywords, and aggregate the scores corresponding to these generations. On the StoryER dataset, CoKe based on our small fine-tuned evaluation models not only reach human-level performance and significantly outperform GPT-4 with a 2x boost in correlation with human annotators, but also requires drastically less # of parameters.
%U https://aclanthology.org/2025.gem-1.31/
%P 366-384
Markdown (Informal)
[CoKe: Customizable Fine-Grained Story Evaluation via Chain-of-Keyword Rationalization](https://aclanthology.org/2025.gem-1.31/) (Joshi et al., GEM 2025)
ACL