@inproceedings{zhang-stadie-2026-leaks,
title = "All Leaks Count, Some Count More: Interpretable Temporal Contamination Detection and Mitigation in {LLM} Backtesting",
author = "Zhang, Zeyu and
Stadie, Bradly",
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.1690/",
doi = "10.18653/2026.findings-emnlp.1690",
pages = "36100--36127",
ISBN = "979-8-89176-462-0",
abstract = "Backtesting LLMs on resolved events assumes models reason only from pre-cutoff knowledge, yet pretrained models inevitably leak post-cutoff knowledge. We introduce a claim-level evaluation framework that decomposes prediction rationales into atomic claims and applies Shapley values to quantify each claim{'}s decision impact, yielding \textbf{Shapley-DCLR} (\textbf{Shapley}-weighted \textbf{D}ecision-\textbf{C}ritical \textbf{L}eakage \textbf{R}ate){---}an interpretable metric measuring what fraction of decision-driving reasoning is contaminated. We further propose \textbf{TimeSPEC} (\textbf{Time}-\textbf{S}upervised \textbf{P}rediction with \textbf{E}xtracted \textbf{C}laims), an inference-time architecture that interleaves temporally-filtered retrieval with claim-level supervision, producing predictions grounded entirely in pre-cutoff evidence. Across three LLMs, the ablation experiments confirm retrieval and supervision are jointly necessary; and a three-task probe further illustrates that the performance cost of temporal enforcement scales with each task{'}s reliance on post-cutoff information."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-stadie-2026-leaks">
<titleInfo>
<title>All Leaks Count, Some Count More: Interpretable Temporal Contamination Detection and Mitigation in LLM Backtesting</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zeyu</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bradly</namePart>
<namePart type="family">Stadie</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>Backtesting LLMs on resolved events assumes models reason only from pre-cutoff knowledge, yet pretrained models inevitably leak post-cutoff knowledge. We introduce a claim-level evaluation framework that decomposes prediction rationales into atomic claims and applies Shapley values to quantify each claim’s decision impact, yielding Shapley-DCLR (Shapley-weighted Decision-Critical Leakage Rate)—an interpretable metric measuring what fraction of decision-driving reasoning is contaminated. We further propose TimeSPEC (Time-Supervised Prediction with Extracted Claims), an inference-time architecture that interleaves temporally-filtered retrieval with claim-level supervision, producing predictions grounded entirely in pre-cutoff evidence. Across three LLMs, the ablation experiments confirm retrieval and supervision are jointly necessary; and a three-task probe further illustrates that the performance cost of temporal enforcement scales with each task’s reliance on post-cutoff information.</abstract>
<identifier type="citekey">zhang-stadie-2026-leaks</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.1690</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.1690/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>36100</start>
<end>36127</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T All Leaks Count, Some Count More: Interpretable Temporal Contamination Detection and Mitigation in LLM Backtesting
%A Zhang, Zeyu
%A Stadie, Bradly
%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 zhang-stadie-2026-leaks
%X Backtesting LLMs on resolved events assumes models reason only from pre-cutoff knowledge, yet pretrained models inevitably leak post-cutoff knowledge. We introduce a claim-level evaluation framework that decomposes prediction rationales into atomic claims and applies Shapley values to quantify each claim’s decision impact, yielding Shapley-DCLR (Shapley-weighted Decision-Critical Leakage Rate)—an interpretable metric measuring what fraction of decision-driving reasoning is contaminated. We further propose TimeSPEC (Time-Supervised Prediction with Extracted Claims), an inference-time architecture that interleaves temporally-filtered retrieval with claim-level supervision, producing predictions grounded entirely in pre-cutoff evidence. Across three LLMs, the ablation experiments confirm retrieval and supervision are jointly necessary; and a three-task probe further illustrates that the performance cost of temporal enforcement scales with each task’s reliance on post-cutoff information.
%R 10.18653/2026.findings-emnlp.1690
%U https://aclanthology.org/2026.findings-emnlp.1690/
%U https://doi.org/10.18653/2026.findings-emnlp.1690
%P 36100-36127
Markdown (Informal)
[All Leaks Count, Some Count More: Interpretable Temporal Contamination Detection and Mitigation in LLM Backtesting](https://aclanthology.org/2026.findings-emnlp.1690/) (Zhang & Stadie, Findings 2026)
ACL