@inproceedings{b-etal-2026-dialpo,
title = "{D}ial{PO}: Dialectical Prompt Optimisation",
author = "B, Gokulakrishnan and
Neeluru, Vijay Kumar and
Puranik, Vinayak S",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.2017/",
doi = "10.18653/2026.findings-emnlp.2017",
pages = "42922--42938",
ISBN = "979-8-89176-462-0",
abstract = "Existing textual-gradient prompt optimisers rely on a single Large Language Model (LLM) to diagnose failures, propose edits, and judge the results, conflating critique and refinement into one model{'}s judgement. We introduce DialPO (Dialectical Prompt Optimisation), a multi-agent framework that replaces the single- model feedback loop with a structured de- bate: an Advocate defends what the prompt does well, a Critic attacks failures, cross- examination filters their arguments adversar- ially, a data-isolated Synthesis model drafts the revised prompt from surviving arguments alone, and a regression gate rejects edits that degrade batch accuracy. We provide a theo- retical framework identifying three precondi- tions under which multi-agent debate improves the feedback-reasoning bottleneck of textual- gradient methods, and show that DialPO in- stantiates them by construction. Across 12 benchmarks spanning classification, reasoning, summarisation, and question answering, Di- alPO achieves the highest scores across various Target models, winning or tying a majority of datasets in head-to-head comparisons against each baseline while using 8{--}18{\texttimes} fewer LLM calls than beam-search optimisers. Compo- nent ablations across five Target models reveal that DialPO{'}s advantage scales inversely with Target-model strength, clarifying when dialec- tical debate helps prompt optimisation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="b-etal-2026-dialpo">
<titleInfo>
<title>DialPO: Dialectical Prompt Optimisation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Gokulakrishnan</namePart>
<namePart type="family">B</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vijay</namePart>
<namePart type="given">Kumar</namePart>
<namePart type="family">Neeluru</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Vinayak</namePart>
<namePart type="given">S</namePart>
<namePart type="family">Puranik</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Existing textual-gradient prompt optimisers rely on a single Large Language Model (LLM) to diagnose failures, propose edits, and judge the results, conflating critique and refinement into one model’s judgement. We introduce DialPO (Dialectical Prompt Optimisation), a multi-agent framework that replaces the single- model feedback loop with a structured de- bate: an Advocate defends what the prompt does well, a Critic attacks failures, cross- examination filters their arguments adversar- ially, a data-isolated Synthesis model drafts the revised prompt from surviving arguments alone, and a regression gate rejects edits that degrade batch accuracy. We provide a theo- retical framework identifying three precondi- tions under which multi-agent debate improves the feedback-reasoning bottleneck of textual- gradient methods, and show that DialPO in- stantiates them by construction. Across 12 benchmarks spanning classification, reasoning, summarisation, and question answering, Di- alPO achieves the highest scores across various Target models, winning or tying a majority of datasets in head-to-head comparisons against each baseline while using 8–18× fewer LLM calls than beam-search optimisers. Compo- nent ablations across five Target models reveal that DialPO’s advantage scales inversely with Target-model strength, clarifying when dialec- tical debate helps prompt optimisation.</abstract>
<identifier type="citekey">b-etal-2026-dialpo</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2017</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2017/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>42922</start>
<end>42938</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DialPO: Dialectical Prompt Optimisation
%A B, Gokulakrishnan
%A Neeluru, Vijay Kumar
%A Puranik, Vinayak S.
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F b-etal-2026-dialpo
%X Existing textual-gradient prompt optimisers rely on a single Large Language Model (LLM) to diagnose failures, propose edits, and judge the results, conflating critique and refinement into one model’s judgement. We introduce DialPO (Dialectical Prompt Optimisation), a multi-agent framework that replaces the single- model feedback loop with a structured de- bate: an Advocate defends what the prompt does well, a Critic attacks failures, cross- examination filters their arguments adversar- ially, a data-isolated Synthesis model drafts the revised prompt from surviving arguments alone, and a regression gate rejects edits that degrade batch accuracy. We provide a theo- retical framework identifying three precondi- tions under which multi-agent debate improves the feedback-reasoning bottleneck of textual- gradient methods, and show that DialPO in- stantiates them by construction. Across 12 benchmarks spanning classification, reasoning, summarisation, and question answering, Di- alPO achieves the highest scores across various Target models, winning or tying a majority of datasets in head-to-head comparisons against each baseline while using 8–18× fewer LLM calls than beam-search optimisers. Compo- nent ablations across five Target models reveal that DialPO’s advantage scales inversely with Target-model strength, clarifying when dialec- tical debate helps prompt optimisation.
%R 10.18653/2026.findings-emnlp.2017
%U https://aclanthology.org/2026.findings-emnlp.2017/
%U https://doi.org/10.18653/2026.findings-emnlp.2017
%P 42922-42938
Markdown (Informal)
[DialPO: Dialectical Prompt Optimisation](https://aclanthology.org/2026.findings-emnlp.2017/) (B et al., Findings 2026)
ACL
- Gokulakrishnan B, Vijay Kumar Neeluru, and Vinayak S Puranik. 2026. DialPO: Dialectical Prompt Optimisation. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 42922–42938, Budapest, Hungary. Association for Computational Linguistics.