@inproceedings{mantro-barrett-2026-ask,
title = "Ask Twice: Bidirectional Cross-Encoder Scoring for Faithfulness Evaluation",
author = "Mantro, Arnav and
Barrett, Leslie",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.24/",
pages = "528--536",
ISBN = "979-8-89176-439-2",
abstract = "Current LLM-based question answering systems provide cited evidence for their claims. However, determining whether each claim within the generated responses is fully supported by its cited evidence is still an ongoing research problem. The task of evaluating attribution quality, or faithfulness to the source text, goes beyond the traditional idea of lexical matching used for relevance in early QA tasks, and the use of lexical models based on semantic distance or word-overlap has declined in favor of NLI-based models and large fine-tuned models. The larger fine-tuned models tend to perform best on current special-purpose benchmarks such as AttributionBench despite struggling with OOD data. Still, exactly why human judgments of faithfulness should align better with LLM evaluators than those based on lexical entailment has never been fully explained. We address the problem with current entailment models by introducing bidirectionality and codifying that idea into a new evaluation modeling approach. We show that this approach is both lightweight and extensible yet competitive with SOTA LLM-based methods, avoiding the pitfalls of both standard NLI and word-overlap approaches. We further show that our model architecture is motivated by a deeper understanding of the gap between lexical distance and faithful claim attribution, exposing the conceptual mechanism behind attribution quality evaluation."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mantro-barrett-2026-ask">
<titleInfo>
<title>Ask Twice: Bidirectional Cross-Encoder Scoring for Faithfulness Evaluation</title>
</titleInfo>
<name type="personal">
<namePart type="given">Arnav</namePart>
<namePart type="family">Mantro</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Leslie</namePart>
<namePart type="family">Barrett</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>Current LLM-based question answering systems provide cited evidence for their claims. However, determining whether each claim within the generated responses is fully supported by its cited evidence is still an ongoing research problem. The task of evaluating attribution quality, or faithfulness to the source text, goes beyond the traditional idea of lexical matching used for relevance in early QA tasks, and the use of lexical models based on semantic distance or word-overlap has declined in favor of NLI-based models and large fine-tuned models. The larger fine-tuned models tend to perform best on current special-purpose benchmarks such as AttributionBench despite struggling with OOD data. Still, exactly why human judgments of faithfulness should align better with LLM evaluators than those based on lexical entailment has never been fully explained. We address the problem with current entailment models by introducing bidirectionality and codifying that idea into a new evaluation modeling approach. We show that this approach is both lightweight and extensible yet competitive with SOTA LLM-based methods, avoiding the pitfalls of both standard NLI and word-overlap approaches. We further show that our model architecture is motivated by a deeper understanding of the gap between lexical distance and faithful claim attribution, exposing the conceptual mechanism behind attribution quality evaluation.</abstract>
<identifier type="citekey">mantro-barrett-2026-ask</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.24/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>528</start>
<end>536</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Ask Twice: Bidirectional Cross-Encoder Scoring for Faithfulness Evaluation
%A Mantro, Arnav
%A Barrett, Leslie
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F mantro-barrett-2026-ask
%X Current LLM-based question answering systems provide cited evidence for their claims. However, determining whether each claim within the generated responses is fully supported by its cited evidence is still an ongoing research problem. The task of evaluating attribution quality, or faithfulness to the source text, goes beyond the traditional idea of lexical matching used for relevance in early QA tasks, and the use of lexical models based on semantic distance or word-overlap has declined in favor of NLI-based models and large fine-tuned models. The larger fine-tuned models tend to perform best on current special-purpose benchmarks such as AttributionBench despite struggling with OOD data. Still, exactly why human judgments of faithfulness should align better with LLM evaluators than those based on lexical entailment has never been fully explained. We address the problem with current entailment models by introducing bidirectionality and codifying that idea into a new evaluation modeling approach. We show that this approach is both lightweight and extensible yet competitive with SOTA LLM-based methods, avoiding the pitfalls of both standard NLI and word-overlap approaches. We further show that our model architecture is motivated by a deeper understanding of the gap between lexical distance and faithful claim attribution, exposing the conceptual mechanism behind attribution quality evaluation.
%U https://aclanthology.org/2026.aacl-long.24/
%P 528-536
Markdown (Informal)
[Ask Twice: Bidirectional Cross-Encoder Scoring for Faithfulness Evaluation](https://aclanthology.org/2026.aacl-long.24/) (Mantro & Barrett, AACL-IJCNLP 2026)
ACL
- Arnav Mantro and Leslie Barrett. 2026. Ask Twice: Bidirectional Cross-Encoder Scoring for Faithfulness Evaluation. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 528–536, Hengqin, China. Association for Computational Linguistics.