@inproceedings{singh-etal-2026-asking,
title = "Asking language models how to represent data for fine-tuning",
author = "Singh, Usneek and
Singha, Ananya and
Awasthi, Abhijeet and
Gulwani, Sumit and
Kanade, Aditya and
Le, Vu and
Singh, Mukul and
Verbruggen, Gust",
editor = "Gupta, Vivek and
Ding, Kaize and
Kokel, Harsha and
Zhao, Yue and
Agarwal, Amit and
Wang, Yu and
Glass, Michael and
Zhang, Yu and
Srinivas, Kavitha and
Chen, Xiusi and
Hassanzadeh, Oktie and
Zhu, Qi and
Chang, Shuaichen and
Luo, Yuan",
booktitle = "Proceedings of the First Workshop on Structured Understanding, Retrieval, and Generation in the {LLM} Era ({SURG}e{LLM} 2026)",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.surgellm-1.13/",
pages = "209--218",
ISBN = "979-8-89176-406-4",
abstract = "Language models are often used for tasks involving structured data like tables and graphs, but there is no principled approach for choosing the best format to represent such data for fine-tuning. We address this in three steps. First, we show that format choice remains important even after fine-tuning; models learn more efficiently with specific formats rather than adapting to any format. Second, we show that a pre-trained model can suggest its own candidate formats by auto-completing partial prompts, reducing reliance on developer intuition. Third, and most importantly, we demonstrate that base model performance across formats reliably predicts post-fine-tuning performance: the format that performs best before fine-tuning remains among the top candidates after fine-tuning in 16 out of 18 settings across three data structure types, three models, and six tasks. This finding allows format selection to be done via inference alone, avoiding costly trial-and-error fine-tuning runs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="singh-etal-2026-asking">
<titleInfo>
<title>Asking language models how to represent data for fine-tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Usneek</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ananya</namePart>
<namePart type="family">Singha</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Abhijeet</namePart>
<namePart type="family">Awasthi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sumit</namePart>
<namePart type="family">Gulwani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aditya</namePart>
<namePart type="family">Kanade</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vu</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mukul</namePart>
<namePart type="family">Singh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Gust</namePart>
<namePart type="family">Verbruggen</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 First Workshop on Structured Understanding, Retrieval, and Generation in the LLM Era (SURGeLLM 2026)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Vivek</namePart>
<namePart type="family">Gupta</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaize</namePart>
<namePart type="family">Ding</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Harsha</namePart>
<namePart type="family">Kokel</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amit</namePart>
<namePart type="family">Agarwal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Michael</namePart>
<namePart type="family">Glass</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kavitha</namePart>
<namePart type="family">Srinivas</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiusi</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Oktie</namePart>
<namePart type="family">Hassanzadeh</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qi</namePart>
<namePart type="family">Zhu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuaichen</namePart>
<namePart type="family">Chang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuan</namePart>
<namePart type="family">Luo</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, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-406-4</identifier>
</relatedItem>
<abstract>Language models are often used for tasks involving structured data like tables and graphs, but there is no principled approach for choosing the best format to represent such data for fine-tuning. We address this in three steps. First, we show that format choice remains important even after fine-tuning; models learn more efficiently with specific formats rather than adapting to any format. Second, we show that a pre-trained model can suggest its own candidate formats by auto-completing partial prompts, reducing reliance on developer intuition. Third, and most importantly, we demonstrate that base model performance across formats reliably predicts post-fine-tuning performance: the format that performs best before fine-tuning remains among the top candidates after fine-tuning in 16 out of 18 settings across three data structure types, three models, and six tasks. This finding allows format selection to be done via inference alone, avoiding costly trial-and-error fine-tuning runs.</abstract>
<identifier type="citekey">singh-etal-2026-asking</identifier>
<location>
<url>https://aclanthology.org/2026.surgellm-1.13/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>209</start>
<end>218</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Asking language models how to represent data for fine-tuning
%A Singh, Usneek
%A Singha, Ananya
%A Awasthi, Abhijeet
%A Gulwani, Sumit
%A Kanade, Aditya
%A Le, Vu
%A Singh, Mukul
%A Verbruggen, Gust
%Y Gupta, Vivek
%Y Ding, Kaize
%Y Kokel, Harsha
%Y Zhao, Yue
%Y Agarwal, Amit
%Y Wang, Yu
%Y Glass, Michael
%Y Zhang, Yu
%Y Srinivas, Kavitha
%Y Chen, Xiusi
%Y Hassanzadeh, Oktie
%Y Zhu, Qi
%Y Chang, Shuaichen
%Y Luo, Yuan
%S Proceedings of the First Workshop on Structured Understanding, Retrieval, and Generation in the LLM Era (SURGeLLM 2026)
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-406-4
%F singh-etal-2026-asking
%X Language models are often used for tasks involving structured data like tables and graphs, but there is no principled approach for choosing the best format to represent such data for fine-tuning. We address this in three steps. First, we show that format choice remains important even after fine-tuning; models learn more efficiently with specific formats rather than adapting to any format. Second, we show that a pre-trained model can suggest its own candidate formats by auto-completing partial prompts, reducing reliance on developer intuition. Third, and most importantly, we demonstrate that base model performance across formats reliably predicts post-fine-tuning performance: the format that performs best before fine-tuning remains among the top candidates after fine-tuning in 16 out of 18 settings across three data structure types, three models, and six tasks. This finding allows format selection to be done via inference alone, avoiding costly trial-and-error fine-tuning runs.
%U https://aclanthology.org/2026.surgellm-1.13/
%P 209-218
Markdown (Informal)
[Asking language models how to represent data for fine-tuning](https://aclanthology.org/2026.surgellm-1.13/) (Singh et al., SURGeLLM 2026)
ACL
- Usneek Singh, Ananya Singha, Abhijeet Awasthi, Sumit Gulwani, Aditya Kanade, Vu Le, Mukul Singh, and Gust Verbruggen. 2026. Asking language models how to represent data for fine-tuning. In Proceedings of the First Workshop on Structured Understanding, Retrieval, and Generation in the LLM Era (SURGeLLM 2026), pages 209–218, San Diego, California, United States. Association for Computational Linguistics.