@inproceedings{liu-etal-2026-capturing,
title = "Capturing Classic Authorial Style in Long-Form Story Generation with {GRPO} Fine-Tuning",
author = "Liu, Jinlong and
Lee, Mark G. and
Bahja, Mohammed and
Kovatchev, Venelin",
editor = "Bonial, Claire and
Berzak, Yevgeni",
booktitle = "Proceedings of the 30th Conference on Computational Natural Language Learning",
month = jul,
year = "2026",
address = "San Diego, California, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.conll-main.31/",
pages = "526--543",
ISBN = "979-8-89176-410-1",
abstract = "Evaluating and optimizing authorial style in long-form story generation is challenging because style judgments often rely on subjective human voting, and there is no stable automatic evaluation method. We propose a two-stage pipeline. First, we train a style-similarity judge by fine-tuning a sentence-transformer with authorship-verification supervision, and calibrate its similarity outputs into a bounded [0,1] reward. Second, we use this judge as the primary reward in Group Relative Policy Optimization (GRPO) to fine-tune an 8B story generator for style-conditioned writing, avoiding the accept/reject supervision required by Direct Preference Optimization (DPO). Across four target authors (Mark Twain, Jane Austen, Charles Dickens, Thomas Hardy), the GRPO-trained 8B model achieves higher style scores than open-weight baselines, with an average style score of 0.893 across authors. These results suggest that AV-calibrated reward modeling provides a practical mechanism for controllable long-form style transfer under moderate model size and training budget."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-capturing">
<titleInfo>
<title>Capturing Classic Authorial Style in Long-Form Story Generation with GRPO Fine-Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jinlong</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mark</namePart>
<namePart type="given">G</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohammed</namePart>
<namePart type="family">Bahja</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Venelin</namePart>
<namePart type="family">Kovatchev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 30th Conference on Computational Natural Language Learning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Claire</namePart>
<namePart type="family">Bonial</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yevgeni</namePart>
<namePart type="family">Berzak</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-410-1</identifier>
</relatedItem>
<abstract>Evaluating and optimizing authorial style in long-form story generation is challenging because style judgments often rely on subjective human voting, and there is no stable automatic evaluation method. We propose a two-stage pipeline. First, we train a style-similarity judge by fine-tuning a sentence-transformer with authorship-verification supervision, and calibrate its similarity outputs into a bounded [0,1] reward. Second, we use this judge as the primary reward in Group Relative Policy Optimization (GRPO) to fine-tune an 8B story generator for style-conditioned writing, avoiding the accept/reject supervision required by Direct Preference Optimization (DPO). Across four target authors (Mark Twain, Jane Austen, Charles Dickens, Thomas Hardy), the GRPO-trained 8B model achieves higher style scores than open-weight baselines, with an average style score of 0.893 across authors. These results suggest that AV-calibrated reward modeling provides a practical mechanism for controllable long-form style transfer under moderate model size and training budget.</abstract>
<identifier type="citekey">liu-etal-2026-capturing</identifier>
<location>
<url>https://aclanthology.org/2026.conll-main.31/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>526</start>
<end>543</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Capturing Classic Authorial Style in Long-Form Story Generation with GRPO Fine-Tuning
%A Liu, Jinlong
%A Lee, Mark G.
%A Bahja, Mohammed
%A Kovatchev, Venelin
%Y Bonial, Claire
%Y Berzak, Yevgeni
%S Proceedings of the 30th Conference on Computational Natural Language Learning
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, USA
%@ 979-8-89176-410-1
%F liu-etal-2026-capturing
%X Evaluating and optimizing authorial style in long-form story generation is challenging because style judgments often rely on subjective human voting, and there is no stable automatic evaluation method. We propose a two-stage pipeline. First, we train a style-similarity judge by fine-tuning a sentence-transformer with authorship-verification supervision, and calibrate its similarity outputs into a bounded [0,1] reward. Second, we use this judge as the primary reward in Group Relative Policy Optimization (GRPO) to fine-tune an 8B story generator for style-conditioned writing, avoiding the accept/reject supervision required by Direct Preference Optimization (DPO). Across four target authors (Mark Twain, Jane Austen, Charles Dickens, Thomas Hardy), the GRPO-trained 8B model achieves higher style scores than open-weight baselines, with an average style score of 0.893 across authors. These results suggest that AV-calibrated reward modeling provides a practical mechanism for controllable long-form style transfer under moderate model size and training budget.
%U https://aclanthology.org/2026.conll-main.31/
%P 526-543
Markdown (Informal)
[Capturing Classic Authorial Style in Long-Form Story Generation with GRPO Fine-Tuning](https://aclanthology.org/2026.conll-main.31/) (Liu et al., CoNLL 2026)
ACL