@inproceedings{huang-etal-2026-much,
title = "How Much Retrieval Depth Does Your Reader Need? Noise Ceilings and Cost-Effectiveness in Table {QA}",
author = "Huang, Jheng-Peng and
Fan, Yao-Chung and
Huang, Jiun-Long",
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.2245/",
doi = "10.18653/2026.findings-emnlp.2245",
pages = "48057--48074",
ISBN = "979-8-89176-462-0",
abstract = "Retrieval and reading in Table QA are typically optimized in isolation, leaving no principled way to balance retrieval precision against reader capacity. We ask: how much retrieval depth does a reader actually need? Across three benchmarks (E2E-WTQ, OTT-QA, and FeTaQA) and four LLMs, we conduct a controlled single-table stress test that guarantees the inclusion of the gold table while increasing retrieval depth to K = 50 using soft or hard distractors. This design isolates reader noise tolerance from retrieval failure. We find that the Noise Ceiling is set jointly by task type, reader, and distractor condition, and is not predictable from reader class alone. With hard negatives, every reader has a Noise Ceiling of K {\ensuremath{\leq}} 5 for long-form answer generation, favoring precision-oriented retrieval. By contrast, several readers remain robust through K = 50 on short-answer and multi-hop QA, depending on the distractor condition. A reader-side Cost per Success analysis further shows that the trade-off between robustness and inference cost varies across readers and distractor conditions. Overall, these results establish the reader{'}s condition-specific Noise Ceiling as a concrete operating target for balancing retrieval depth against precision in controlled single-table QA."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-much">
<titleInfo>
<title>How Much Retrieval Depth Does Your Reader Need? Noise Ceilings and Cost-Effectiveness in Table QA</title>
</titleInfo>
<name type="personal">
<namePart type="given">Jheng-Peng</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yao-Chung</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiun-Long</namePart>
<namePart type="family">Huang</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>Retrieval and reading in Table QA are typically optimized in isolation, leaving no principled way to balance retrieval precision against reader capacity. We ask: how much retrieval depth does a reader actually need? Across three benchmarks (E2E-WTQ, OTT-QA, and FeTaQA) and four LLMs, we conduct a controlled single-table stress test that guarantees the inclusion of the gold table while increasing retrieval depth to K = 50 using soft or hard distractors. This design isolates reader noise tolerance from retrieval failure. We find that the Noise Ceiling is set jointly by task type, reader, and distractor condition, and is not predictable from reader class alone. With hard negatives, every reader has a Noise Ceiling of K \ensuremathłeq 5 for long-form answer generation, favoring precision-oriented retrieval. By contrast, several readers remain robust through K = 50 on short-answer and multi-hop QA, depending on the distractor condition. A reader-side Cost per Success analysis further shows that the trade-off between robustness and inference cost varies across readers and distractor conditions. Overall, these results establish the reader’s condition-specific Noise Ceiling as a concrete operating target for balancing retrieval depth against precision in controlled single-table QA.</abstract>
<identifier type="citekey">huang-etal-2026-much</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2245</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2245/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>48057</start>
<end>48074</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T How Much Retrieval Depth Does Your Reader Need? Noise Ceilings and Cost-Effectiveness in Table QA
%A Huang, Jheng-Peng
%A Fan, Yao-Chung
%A Huang, Jiun-Long
%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 huang-etal-2026-much
%X Retrieval and reading in Table QA are typically optimized in isolation, leaving no principled way to balance retrieval precision against reader capacity. We ask: how much retrieval depth does a reader actually need? Across three benchmarks (E2E-WTQ, OTT-QA, and FeTaQA) and four LLMs, we conduct a controlled single-table stress test that guarantees the inclusion of the gold table while increasing retrieval depth to K = 50 using soft or hard distractors. This design isolates reader noise tolerance from retrieval failure. We find that the Noise Ceiling is set jointly by task type, reader, and distractor condition, and is not predictable from reader class alone. With hard negatives, every reader has a Noise Ceiling of K \ensuremathłeq 5 for long-form answer generation, favoring precision-oriented retrieval. By contrast, several readers remain robust through K = 50 on short-answer and multi-hop QA, depending on the distractor condition. A reader-side Cost per Success analysis further shows that the trade-off between robustness and inference cost varies across readers and distractor conditions. Overall, these results establish the reader’s condition-specific Noise Ceiling as a concrete operating target for balancing retrieval depth against precision in controlled single-table QA.
%R 10.18653/2026.findings-emnlp.2245
%U https://aclanthology.org/2026.findings-emnlp.2245/
%U https://doi.org/10.18653/2026.findings-emnlp.2245
%P 48057-48074
Markdown (Informal)
[How Much Retrieval Depth Does Your Reader Need? Noise Ceilings and Cost-Effectiveness in Table QA](https://aclanthology.org/2026.findings-emnlp.2245/) (Huang et al., Findings 2026)
ACL