@inproceedings{sun-etal-2026-bear,
title = "{BEAR}: Budgeted Evidence Allocation for Multi-Document Reasoning",
author = "Sun, Lin and
Zhang, Linglin and
Huang, Jingang and
Jia, Change and
Cheng, Zhengwei and
Zhang, Xiangzheng",
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.1348/",
pages = "30731--30749",
ISBN = "979-8-89176-459-0",
abstract = "We argue that multi-document reasoning is constrained not only by how much text a model can read, but also by how a limited query-time evidence budget is allocated across documents and semantic granularities. Full-context inference exposes the model to broad evidence non-selectively and at high per-query cost, while flat chunk retrieval often returns locally relevant passages that are weakly organized for cross-document synthesis. We present \textbf{BEAR}, a framework for structured evidence allocation that builds hierarchical semantic indices offline and performs coarse-to-fine evidence access at query time through complementary \textit{exploration} and \textit{recovery} paths. This design formulates coarse-to-fine reasoning as structured evidence allocation under a fixed final evidence-context budget. Across nine evaluation settings spanning synthetic evidence allocation, multi-hop QA, agentic retrieval, conversational memory, long-document QA, and open-domain QA, BEAR delivers strong coverage{--}faithfulness trade-offs on DragonBall, achieves the best retrieval-based results on 2WikiMultiHopQA, LoCoMo, QASPER, MuSiQue, and PopQA, and remains competitive on HotpotQA and NarrativeQA. Under the matched retrieval protocol, BEAR operates with an approximately 4K-token final evidence context, substantially smaller than the inputs used by the long-context references. These results show that evidence exposure{---}not context length alone{---}should be selectively organized, with gains arising from hierarchy as an allocation substrate and complementary exploration and recovery rather than semantic chunking alone."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="sun-etal-2026-bear">
<titleInfo>
<title>BEAR: Budgeted Evidence Allocation for Multi-Document Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Lin</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Linglin</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jingang</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Change</namePart>
<namePart type="family">Jia</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhengwei</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiangzheng</namePart>
<namePart type="family">Zhang</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>We argue that multi-document reasoning is constrained not only by how much text a model can read, but also by how a limited query-time evidence budget is allocated across documents and semantic granularities. Full-context inference exposes the model to broad evidence non-selectively and at high per-query cost, while flat chunk retrieval often returns locally relevant passages that are weakly organized for cross-document synthesis. We present BEAR, a framework for structured evidence allocation that builds hierarchical semantic indices offline and performs coarse-to-fine evidence access at query time through complementary exploration and recovery paths. This design formulates coarse-to-fine reasoning as structured evidence allocation under a fixed final evidence-context budget. Across nine evaluation settings spanning synthetic evidence allocation, multi-hop QA, agentic retrieval, conversational memory, long-document QA, and open-domain QA, BEAR delivers strong coverage–faithfulness trade-offs on DragonBall, achieves the best retrieval-based results on 2WikiMultiHopQA, LoCoMo, QASPER, MuSiQue, and PopQA, and remains competitive on HotpotQA and NarrativeQA. Under the matched retrieval protocol, BEAR operates with an approximately 4K-token final evidence context, substantially smaller than the inputs used by the long-context references. These results show that evidence exposure—not context length alone—should be selectively organized, with gains arising from hierarchy as an allocation substrate and complementary exploration and recovery rather than semantic chunking alone.</abstract>
<identifier type="citekey">sun-etal-2026-bear</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1348/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>30731</start>
<end>30749</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T BEAR: Budgeted Evidence Allocation for Multi-Document Reasoning
%A Sun, Lin
%A Zhang, Linglin
%A Huang, Jingang
%A Jia, Change
%A Cheng, Zhengwei
%A Zhang, Xiangzheng
%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 sun-etal-2026-bear
%X We argue that multi-document reasoning is constrained not only by how much text a model can read, but also by how a limited query-time evidence budget is allocated across documents and semantic granularities. Full-context inference exposes the model to broad evidence non-selectively and at high per-query cost, while flat chunk retrieval often returns locally relevant passages that are weakly organized for cross-document synthesis. We present BEAR, a framework for structured evidence allocation that builds hierarchical semantic indices offline and performs coarse-to-fine evidence access at query time through complementary exploration and recovery paths. This design formulates coarse-to-fine reasoning as structured evidence allocation under a fixed final evidence-context budget. Across nine evaluation settings spanning synthetic evidence allocation, multi-hop QA, agentic retrieval, conversational memory, long-document QA, and open-domain QA, BEAR delivers strong coverage–faithfulness trade-offs on DragonBall, achieves the best retrieval-based results on 2WikiMultiHopQA, LoCoMo, QASPER, MuSiQue, and PopQA, and remains competitive on HotpotQA and NarrativeQA. Under the matched retrieval protocol, BEAR operates with an approximately 4K-token final evidence context, substantially smaller than the inputs used by the long-context references. These results show that evidence exposure—not context length alone—should be selectively organized, with gains arising from hierarchy as an allocation substrate and complementary exploration and recovery rather than semantic chunking alone.
%U https://aclanthology.org/2026.emnlp-main.1348/
%P 30731-30749
Markdown (Informal)
[BEAR: Budgeted Evidence Allocation for Multi-Document Reasoning](https://aclanthology.org/2026.emnlp-main.1348/) (Sun et al., EMNLP 2026)
ACL
- Lin Sun, Linglin Zhang, Jingang Huang, Change Jia, Zhengwei Cheng, and Xiangzheng Zhang. 2026. BEAR: Budgeted Evidence Allocation for Multi-Document Reasoning. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 30731–30749, Budapest, Hungary. Association for Computational Linguistics.