@inproceedings{halil-etal-2026-gnosis,
title = "Gnosis Deepsearch: Structured Search Space Navigation for Complex {QA}",
author = "Halil, Umut and
Zhang, Le and
Huang, Jin and
Zhang, Yuwei and
Hou, Xinyu and
Melo, Andre and
Piramanayagam, Shriram and
Fejza, Amela and
Lai, Ruofei and
Ren, Yang and
Vougiouklis, Pavlos and
Pan, Jeff Z.",
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.2164/",
pages = "46292--46315",
ISBN = "979-8-89176-462-0",
abstract = "Complex, multi-step reasoning demands agents to be capable of dynamically exploring diverse research directions and adapting plans on the fly. We introduce Gnosis Deepsearch, a schema-guided framework for long-horizon planning in complex Question Answering, where variables, clues, and constraints are captured in a structured state representation maintained across trajectories. This state is cascaded into parallel search paths broadening exploration without relying on sequential execution, thereby mitigating both limited search space coverage and catastrophic forgetting. We couple Gnosis with a retriever that sidesteps challenges associated with retrieval and LLM reading when longer documents are involved. Our retrieval strategy ensures that the most relevant snippets from longer documents are considered, showcasing several benefits within both Gnosis and a more general ReAct framework on the BrowseComp-Plus benchmark. Extensive experiments on several QA benchmarks of varying difficulty demonstrate the strong performance of Gnosis across different configurations."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="halil-etal-2026-gnosis">
<titleInfo>
<title>Gnosis Deepsearch: Structured Search Space Navigation for Complex QA</title>
</titleInfo>
<name type="personal">
<namePart type="given">Umut</namePart>
<namePart type="family">Halil</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Le</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jin</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuwei</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinyu</namePart>
<namePart type="family">Hou</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andre</namePart>
<namePart type="family">Melo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shriram</namePart>
<namePart type="family">Piramanayagam</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Amela</namePart>
<namePart type="family">Fejza</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruofei</namePart>
<namePart type="family">Lai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Ren</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Pavlos</namePart>
<namePart type="family">Vougiouklis</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeff</namePart>
<namePart type="given">Z</namePart>
<namePart type="family">Pan</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>Complex, multi-step reasoning demands agents to be capable of dynamically exploring diverse research directions and adapting plans on the fly. We introduce Gnosis Deepsearch, a schema-guided framework for long-horizon planning in complex Question Answering, where variables, clues, and constraints are captured in a structured state representation maintained across trajectories. This state is cascaded into parallel search paths broadening exploration without relying on sequential execution, thereby mitigating both limited search space coverage and catastrophic forgetting. We couple Gnosis with a retriever that sidesteps challenges associated with retrieval and LLM reading when longer documents are involved. Our retrieval strategy ensures that the most relevant snippets from longer documents are considered, showcasing several benefits within both Gnosis and a more general ReAct framework on the BrowseComp-Plus benchmark. Extensive experiments on several QA benchmarks of varying difficulty demonstrate the strong performance of Gnosis across different configurations.</abstract>
<identifier type="citekey">halil-etal-2026-gnosis</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2164/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>46292</start>
<end>46315</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Gnosis Deepsearch: Structured Search Space Navigation for Complex QA
%A Halil, Umut
%A Zhang, Le
%A Huang, Jin
%A Zhang, Yuwei
%A Hou, Xinyu
%A Melo, Andre
%A Piramanayagam, Shriram
%A Fejza, Amela
%A Lai, Ruofei
%A Ren, Yang
%A Vougiouklis, Pavlos
%A Pan, Jeff Z.
%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 halil-etal-2026-gnosis
%X Complex, multi-step reasoning demands agents to be capable of dynamically exploring diverse research directions and adapting plans on the fly. We introduce Gnosis Deepsearch, a schema-guided framework for long-horizon planning in complex Question Answering, where variables, clues, and constraints are captured in a structured state representation maintained across trajectories. This state is cascaded into parallel search paths broadening exploration without relying on sequential execution, thereby mitigating both limited search space coverage and catastrophic forgetting. We couple Gnosis with a retriever that sidesteps challenges associated with retrieval and LLM reading when longer documents are involved. Our retrieval strategy ensures that the most relevant snippets from longer documents are considered, showcasing several benefits within both Gnosis and a more general ReAct framework on the BrowseComp-Plus benchmark. Extensive experiments on several QA benchmarks of varying difficulty demonstrate the strong performance of Gnosis across different configurations.
%U https://aclanthology.org/2026.findings-emnlp.2164/
%P 46292-46315
Markdown (Informal)
[Gnosis Deepsearch: Structured Search Space Navigation for Complex QA](https://aclanthology.org/2026.findings-emnlp.2164/) (Halil et al., Findings 2026)
ACL
- Umut Halil, Le Zhang, Jin Huang, Yuwei Zhang, Xinyu Hou, Andre Melo, Shriram Piramanayagam, Amela Fejza, Ruofei Lai, Yang Ren, Pavlos Vougiouklis, and Jeff Z. Pan. 2026. Gnosis Deepsearch: Structured Search Space Navigation for Complex QA. In Findings of the Association for Computational Linguistics: EMNLP 2026, pages 46292–46315, Budapest, Hungary. Association for Computational Linguistics.