@inproceedings{liang-etal-2026-probetqa,
title = "{P}robe{TQA}: Verifiable Table Question Answering via Multi-Round Executable Probing",
author = "Liang, Yan and
Ren, Huimin and
Zhang, Xindi and
Xin, Hongsheng and
Huo, Wenbin and
Guo, Yanan and
Zheng, Jinpeng and
Zhang, Kaike and
Zhan, Kun and
Hu, Jiani",
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.900/",
pages = "20715--20741",
ISBN = "979-8-89176-459-0",
abstract = "Enabling Large Language Models (LLMs) to perform tool-free Table Question Answering (TableQA) presents two critical challenges: ``question-first'' data synthesis often introduces logical hallucinations, while standard reinforcement learning (RL) suffers from sparse rewards over long reasoning chains. We propose ProbeTQA, a unified framework that couples execution-verified data construction with process-supervised RL. To mitigate synthesis hallucinations, ProbeTQA adopts an answer-first executable probing paradigm, in which an agent performs multi-round exploratory programming over real tables so that every synthesized instance is grounded in a verifiable witness program. To address reward sparsity, we convert verified code trajectories into natural-language reasoning traces and derive execution-grounded process rubrics directly from runtime states, providing reliable dense supervision. Using these rubrics, we optimize the model with Group Relative Policy Optimization (GRPO) under a Two-Stage Reward Curriculum, thereby internalizing complex tool-use reasoning into the model parameters. Experiments show that our 8B model achieves state-of-the-art tool-free performance on challenging benchmarks, attaining 65.8{\%} accuracy on TableBench and outperforming GPT-5.4 and DeepSeek-v3.2 by 13.4 and 6.8 percentage points, respectively."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liang-etal-2026-probetqa">
<titleInfo>
<title>ProbeTQA: Verifiable Table Question Answering via Multi-Round Executable Probing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yan</namePart>
<namePart type="family">Liang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Huimin</namePart>
<namePart type="family">Ren</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xindi</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hongsheng</namePart>
<namePart type="family">Xin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenbin</namePart>
<namePart type="family">Huo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yanan</namePart>
<namePart type="family">Guo</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinpeng</namePart>
<namePart type="family">Zheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaike</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kun</namePart>
<namePart type="family">Zhan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiani</namePart>
<namePart type="family">Hu</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>Enabling Large Language Models (LLMs) to perform tool-free Table Question Answering (TableQA) presents two critical challenges: “question-first” data synthesis often introduces logical hallucinations, while standard reinforcement learning (RL) suffers from sparse rewards over long reasoning chains. We propose ProbeTQA, a unified framework that couples execution-verified data construction with process-supervised RL. To mitigate synthesis hallucinations, ProbeTQA adopts an answer-first executable probing paradigm, in which an agent performs multi-round exploratory programming over real tables so that every synthesized instance is grounded in a verifiable witness program. To address reward sparsity, we convert verified code trajectories into natural-language reasoning traces and derive execution-grounded process rubrics directly from runtime states, providing reliable dense supervision. Using these rubrics, we optimize the model with Group Relative Policy Optimization (GRPO) under a Two-Stage Reward Curriculum, thereby internalizing complex tool-use reasoning into the model parameters. Experiments show that our 8B model achieves state-of-the-art tool-free performance on challenging benchmarks, attaining 65.8% accuracy on TableBench and outperforming GPT-5.4 and DeepSeek-v3.2 by 13.4 and 6.8 percentage points, respectively.</abstract>
<identifier type="citekey">liang-etal-2026-probetqa</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.900/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>20715</start>
<end>20741</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ProbeTQA: Verifiable Table Question Answering via Multi-Round Executable Probing
%A Liang, Yan
%A Ren, Huimin
%A Zhang, Xindi
%A Xin, Hongsheng
%A Huo, Wenbin
%A Guo, Yanan
%A Zheng, Jinpeng
%A Zhang, Kaike
%A Zhan, Kun
%A Hu, Jiani
%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 liang-etal-2026-probetqa
%X Enabling Large Language Models (LLMs) to perform tool-free Table Question Answering (TableQA) presents two critical challenges: “question-first” data synthesis often introduces logical hallucinations, while standard reinforcement learning (RL) suffers from sparse rewards over long reasoning chains. We propose ProbeTQA, a unified framework that couples execution-verified data construction with process-supervised RL. To mitigate synthesis hallucinations, ProbeTQA adopts an answer-first executable probing paradigm, in which an agent performs multi-round exploratory programming over real tables so that every synthesized instance is grounded in a verifiable witness program. To address reward sparsity, we convert verified code trajectories into natural-language reasoning traces and derive execution-grounded process rubrics directly from runtime states, providing reliable dense supervision. Using these rubrics, we optimize the model with Group Relative Policy Optimization (GRPO) under a Two-Stage Reward Curriculum, thereby internalizing complex tool-use reasoning into the model parameters. Experiments show that our 8B model achieves state-of-the-art tool-free performance on challenging benchmarks, attaining 65.8% accuracy on TableBench and outperforming GPT-5.4 and DeepSeek-v3.2 by 13.4 and 6.8 percentage points, respectively.
%U https://aclanthology.org/2026.emnlp-main.900/
%P 20715-20741
Markdown (Informal)
[ProbeTQA: Verifiable Table Question Answering via Multi-Round Executable Probing](https://aclanthology.org/2026.emnlp-main.900/) (Liang et al., EMNLP 2026)
ACL
- Yan Liang, Huimin Ren, Xindi Zhang, Hongsheng Xin, Wenbin Huo, Yanan Guo, Jinpeng Zheng, Kaike Zhang, Kun Zhan, and Jiani Hu. 2026. ProbeTQA: Verifiable Table Question Answering via Multi-Round Executable Probing. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 20715–20741, Budapest, Hungary. Association for Computational Linguistics.