@inproceedings{rodriguez-etal-2026-improving,
title = "Improving {LLM}s via Validator-to-Generator Alignment",
author = "Rodriguez, Juan Diego and
Zhang, Jocelyn and
Erk, Katrin and
Durrett, Greg",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2411/",
pages = "54769--54786",
ISBN = "979-8-89176-459-0",
abstract = "Large language models are inconsistent: varying prompts or including unrelated information can lead to unexpected changes in model outputs. The generator-validator (G-V) gap is one manifestation of this phenomenon, where LLMs generate responses that they then deem as invalid if re-queried to validate them. In this work, we introduce a new formulation of G-V consistency that involves a principled correction for utterance frequency. Specifically, generators often assign low likelihood to valid strings simply because those strings are a priori unlikely, which makes naive notions of G-V consistency unworkable. We show that under a natural model of rational agents answering questions with multiple answers, consistency of the validator with a frequency-corrected generator score emerges naturally. Our method, Frequency-corrected Learning of Ordered Rank Alignment (FLORA), is a training objective implementing frequency-corrected G-V consistency for real-world LLMs. Our experimental results show that training with FLORA substantially improves both G-V consistency and generator performance over prior methods, with gains of up to +27pp in Pearson correlation on IFEval and HumanEval, while preserving validator quality across all evaluated task"
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="rodriguez-etal-2026-improving">
<titleInfo>
<title>Improving LLMs via Validator-to-Generator Alignment</title>
</titleInfo>
<name type="personal">
<namePart type="given">Juan</namePart>
<namePart type="given">Diego</namePart>
<namePart type="family">Rodriguez</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jocelyn</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Katrin</namePart>
<namePart type="family">Erk</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Greg</namePart>
<namePart type="family">Durrett</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Large language models are inconsistent: varying prompts or including unrelated information can lead to unexpected changes in model outputs. The generator-validator (G-V) gap is one manifestation of this phenomenon, where LLMs generate responses that they then deem as invalid if re-queried to validate them. In this work, we introduce a new formulation of G-V consistency that involves a principled correction for utterance frequency. Specifically, generators often assign low likelihood to valid strings simply because those strings are a priori unlikely, which makes naive notions of G-V consistency unworkable. We show that under a natural model of rational agents answering questions with multiple answers, consistency of the validator with a frequency-corrected generator score emerges naturally. Our method, Frequency-corrected Learning of Ordered Rank Alignment (FLORA), is a training objective implementing frequency-corrected G-V consistency for real-world LLMs. Our experimental results show that training with FLORA substantially improves both G-V consistency and generator performance over prior methods, with gains of up to +27pp in Pearson correlation on IFEval and HumanEval, while preserving validator quality across all evaluated task</abstract>
<identifier type="citekey">rodriguez-etal-2026-improving</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2411/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>54769</start>
<end>54786</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Improving LLMs via Validator-to-Generator Alignment
%A Rodriguez, Juan Diego
%A Zhang, Jocelyn
%A Erk, Katrin
%A Durrett, Greg
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F rodriguez-etal-2026-improving
%X Large language models are inconsistent: varying prompts or including unrelated information can lead to unexpected changes in model outputs. The generator-validator (G-V) gap is one manifestation of this phenomenon, where LLMs generate responses that they then deem as invalid if re-queried to validate them. In this work, we introduce a new formulation of G-V consistency that involves a principled correction for utterance frequency. Specifically, generators often assign low likelihood to valid strings simply because those strings are a priori unlikely, which makes naive notions of G-V consistency unworkable. We show that under a natural model of rational agents answering questions with multiple answers, consistency of the validator with a frequency-corrected generator score emerges naturally. Our method, Frequency-corrected Learning of Ordered Rank Alignment (FLORA), is a training objective implementing frequency-corrected G-V consistency for real-world LLMs. Our experimental results show that training with FLORA substantially improves both G-V consistency and generator performance over prior methods, with gains of up to +27pp in Pearson correlation on IFEval and HumanEval, while preserving validator quality across all evaluated task
%U https://aclanthology.org/2026.emnlp-main.2411/
%P 54769-54786
Markdown (Informal)
[Improving LLMs via Validator-to-Generator Alignment](https://aclanthology.org/2026.emnlp-main.2411/) (Rodriguez et al., EMNLP 2026)
ACL
- Juan Diego Rodriguez, Jocelyn Zhang, Katrin Erk, and Greg Durrett. 2026. Improving LLMs via Validator-to-Generator Alignment. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 54769–54786, Budapest, Hungary. Association for Computational Linguistics.