@inproceedings{yao-etal-2026-correct,
title = "When the Correct Law Still Fails: A Diagnostic Benchmark for Law-in-Force Retrieval in Versioned Legal {RAG}",
author = "Yao, Yuchen and
HU, Yiran and
Zhang, Kaixiang and
Su, Yixin and
Ge, Jiani and
Liu, Yun",
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.42/",
pages = "879--893",
ISBN = "979-8-89176-439-2",
abstract = "Legal RAG systems often retrieve passages by semantic relevance, lexical score, or citation match. In versioned legal corpora, however, a passage can be topically relevant and citation-matched while being legally invalid for the operative date. We study law-in-force retrieval: retrieving evidence that matches the authority, is valid at the operative date, and supports the answer. We build LawInForce-eCFR from 805 human-validated eCFR/CFR amendment pairs and 3,220 controlled answer-flipping yes/no questions. We also present AuthorityFirst-RAG, a transparent diagnostic retriever that decomposes retrieval into authority identification, temporal scoping, and rule-support ranking. AuthorityFirst-RAG improves Recall@1 from 2.80{\%} for BM25-date-filter to 30.90{\%}, but generation remains difficult even with GoldOracle evidence. These results show a dual bottleneck in versioned legal RAG: retrieving the law in force and applying numeric legal rules."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="yao-etal-2026-correct">
<titleInfo>
<title>When the Correct Law Still Fails: A Diagnostic Benchmark for Law-in-Force Retrieval in Versioned Legal RAG</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuchen</namePart>
<namePart type="family">Yao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yiran</namePart>
<namePart type="family">HU</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kaixiang</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yixin</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiani</namePart>
<namePart type="family">Ge</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun</namePart>
<namePart type="family">Liu</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>Legal RAG systems often retrieve passages by semantic relevance, lexical score, or citation match. In versioned legal corpora, however, a passage can be topically relevant and citation-matched while being legally invalid for the operative date. We study law-in-force retrieval: retrieving evidence that matches the authority, is valid at the operative date, and supports the answer. We build LawInForce-eCFR from 805 human-validated eCFR/CFR amendment pairs and 3,220 controlled answer-flipping yes/no questions. We also present AuthorityFirst-RAG, a transparent diagnostic retriever that decomposes retrieval into authority identification, temporal scoping, and rule-support ranking. AuthorityFirst-RAG improves Recall@1 from 2.80% for BM25-date-filter to 30.90%, but generation remains difficult even with GoldOracle evidence. These results show a dual bottleneck in versioned legal RAG: retrieving the law in force and applying numeric legal rules.</abstract>
<identifier type="citekey">yao-etal-2026-correct</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.42/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>879</start>
<end>893</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T When the Correct Law Still Fails: A Diagnostic Benchmark for Law-in-Force Retrieval in Versioned Legal RAG
%A Yao, Yuchen
%A HU, Yiran
%A Zhang, Kaixiang
%A Su, Yixin
%A Ge, Jiani
%A Liu, Yun
%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 yao-etal-2026-correct
%X Legal RAG systems often retrieve passages by semantic relevance, lexical score, or citation match. In versioned legal corpora, however, a passage can be topically relevant and citation-matched while being legally invalid for the operative date. We study law-in-force retrieval: retrieving evidence that matches the authority, is valid at the operative date, and supports the answer. We build LawInForce-eCFR from 805 human-validated eCFR/CFR amendment pairs and 3,220 controlled answer-flipping yes/no questions. We also present AuthorityFirst-RAG, a transparent diagnostic retriever that decomposes retrieval into authority identification, temporal scoping, and rule-support ranking. AuthorityFirst-RAG improves Recall@1 from 2.80% for BM25-date-filter to 30.90%, but generation remains difficult even with GoldOracle evidence. These results show a dual bottleneck in versioned legal RAG: retrieving the law in force and applying numeric legal rules.
%U https://aclanthology.org/2026.aacl-long.42/
%P 879-893
Markdown (Informal)
[When the Correct Law Still Fails: A Diagnostic Benchmark for Law-in-Force Retrieval in Versioned Legal RAG](https://aclanthology.org/2026.aacl-long.42/) (Yao et al., AACL-IJCNLP 2026)
ACL