@inproceedings{sun-etal-2026-aegis,
title = "{AEGIS}-{D}: Multi-Agent Debate for High-Precision Verification of Hard Competition Mathematics",
author = "Sun, Jiayang and
Xu, Jiawei and
Pan, Maxm and
Cheng, Zerui and
Viswanath, Pramod",
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.890/",
pages = "18861--18873",
ISBN = "979-8-89176-462-0",
abstract = "Fluent but subtly wrong mathematical solutions can contaminate training, evaluation, and reward-labeling data. We present AEGIS-D (Assessment- and Evidence-Gated Inspection System via Debate), a precision-first multi-agent pipeline for blind verification of competition-level mathematics. Five same-model, role-prompted agents exchange structured arguments, while an assessment gate accepts a candidate only when at least three report positive confirming evidence rather than mere non-refutation. On 195 hard-to-verify variants from 58 IMO/USAMO/Putnam-level problems, a single judge achieves 55.1{\%} precision; AEGIS-D achieves 92.5{\%} precision and 63.8{\%} recall, reducing false positives from 24.7 to 3.0 per run (8.2-fold). A repeated two-by-two factorial associates role prompting with the largest observed false-positive reduction and finds a smaller conditional benefit from gating. Paired round and threshold analyses expose the cost-precision trade-off, and single-run transfers to Omni-MATH and OlymMATH provide preliminary evidence beyond the constructed stress test."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sun-etal-2026-aegis">
<titleInfo>
<title>AEGIS-D: Multi-Agent Debate for High-Precision Verification of Hard Competition Mathematics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jiayang</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiawei</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Maxm</namePart>
<namePart type="family">Pan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zerui</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pramod</namePart>
<namePart type="family">Viswanath</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>Fluent but subtly wrong mathematical solutions can contaminate training, evaluation, and reward-labeling data. We present AEGIS-D (Assessment- and Evidence-Gated Inspection System via Debate), a precision-first multi-agent pipeline for blind verification of competition-level mathematics. Five same-model, role-prompted agents exchange structured arguments, while an assessment gate accepts a candidate only when at least three report positive confirming evidence rather than mere non-refutation. On 195 hard-to-verify variants from 58 IMO/USAMO/Putnam-level problems, a single judge achieves 55.1% precision; AEGIS-D achieves 92.5% precision and 63.8% recall, reducing false positives from 24.7 to 3.0 per run (8.2-fold). A repeated two-by-two factorial associates role prompting with the largest observed false-positive reduction and finds a smaller conditional benefit from gating. Paired round and threshold analyses expose the cost-precision trade-off, and single-run transfers to Omni-MATH and OlymMATH provide preliminary evidence beyond the constructed stress test.</abstract>
<identifier type="citekey">sun-etal-2026-aegis</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.890/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18861</start>
<end>18873</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AEGIS-D: Multi-Agent Debate for High-Precision Verification of Hard Competition Mathematics
%A Sun, Jiayang
%A Xu, Jiawei
%A Pan, Maxm
%A Cheng, Zerui
%A Viswanath, Pramod
%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 sun-etal-2026-aegis
%X Fluent but subtly wrong mathematical solutions can contaminate training, evaluation, and reward-labeling data. We present AEGIS-D (Assessment- and Evidence-Gated Inspection System via Debate), a precision-first multi-agent pipeline for blind verification of competition-level mathematics. Five same-model, role-prompted agents exchange structured arguments, while an assessment gate accepts a candidate only when at least three report positive confirming evidence rather than mere non-refutation. On 195 hard-to-verify variants from 58 IMO/USAMO/Putnam-level problems, a single judge achieves 55.1% precision; AEGIS-D achieves 92.5% precision and 63.8% recall, reducing false positives from 24.7 to 3.0 per run (8.2-fold). A repeated two-by-two factorial associates role prompting with the largest observed false-positive reduction and finds a smaller conditional benefit from gating. Paired round and threshold analyses expose the cost-precision trade-off, and single-run transfers to Omni-MATH and OlymMATH provide preliminary evidence beyond the constructed stress test.
%U https://aclanthology.org/2026.findings-emnlp.890/
%P 18861-18873
Markdown (Informal)
[AEGIS-D: Multi-Agent Debate for High-Precision Verification of Hard Competition Mathematics](https://aclanthology.org/2026.findings-emnlp.890/) (Sun et al., Findings 2026)
ACL