@inproceedings{ram-etal-2026-validity,
title = "The Validity{--}Fidelity Gap: A Controlled Study of Representation Drift in Structured Extraction Under Redundant Contexts",
author = "Ram, Gokul Srinath Seetha and
Vasudevan, Anirudh and
Elavazhagan, Rashmi",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.107/",
pages = "2112--2126",
ISBN = "979-8-89176-444-6",
abstract = "Pipelines that use LLMs for structured extraction are commonly monitored by checking that each output parses as JSON and matches the expected keys. We show that this check can stay perfect while the extracted content degrades. We call the failure representation drift: the model still finds an entity but writes it in a different form than the intended one, for example a nested key-value object where a plain string is expected. The output is still valid JSON, yet the entity is scored both as missing (a false negative) and as an unexpected extra (a false positive). We study this with synthetic clinical notes repeated verbatim 1x-32x, which gives exact ground truth with prompt, schema and decoding held fixed. For LLaMA-3.1-8B-Instruct, micro-F1 falls from 0.499 to 0.259 while JSON validity stays 1.00. All 280 medication entities scored as missing in a 100-patient run are present in the output as nested objects, and a follow-up request with no schema recovers 96.7{\%} of medications versus 40.0{\%} under the schema at 16x. We introduce the Differential Redundancy Effect (DRE), the extra F1 drop under repeated content relative to an approximately length-matched unrelated-filler control, and use it as a diagnostic contrast rather than a causal estimate. The effect is scale-dependent: DRE is large at 8B and near zero at 70B and Qwen2.5-72B, where F1 does not degrade. Validity metrics stay at 1.00 in both regimes, so they carry no information about content fidelity. A smaller synthetic news experiment shows the same dissociation with a different schema. We also give a zero-ground-truth drift check and monitoring recommendations. All evidence is synthetic, and we make no claim about real clinical notes."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="ram-etal-2026-validity">
<titleInfo>
<title>The Validity–Fidelity Gap: A Controlled Study of Representation Drift in Structured Extraction Under Redundant Contexts</title>
</titleInfo>
<name type="personal">
<namePart type="given">Gokul</namePart>
<namePart type="given">Srinath</namePart>
<namePart type="given">Seetha</namePart>
<namePart type="family">Ram</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anirudh</namePart>
<namePart type="family">Vasudevan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Rashmi</namePart>
<namePart type="family">Elavazhagan</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>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</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">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Pipelines that use LLMs for structured extraction are commonly monitored by checking that each output parses as JSON and matches the expected keys. We show that this check can stay perfect while the extracted content degrades. We call the failure representation drift: the model still finds an entity but writes it in a different form than the intended one, for example a nested key-value object where a plain string is expected. The output is still valid JSON, yet the entity is scored both as missing (a false negative) and as an unexpected extra (a false positive). We study this with synthetic clinical notes repeated verbatim 1x-32x, which gives exact ground truth with prompt, schema and decoding held fixed. For LLaMA-3.1-8B-Instruct, micro-F1 falls from 0.499 to 0.259 while JSON validity stays 1.00. All 280 medication entities scored as missing in a 100-patient run are present in the output as nested objects, and a follow-up request with no schema recovers 96.7% of medications versus 40.0% under the schema at 16x. We introduce the Differential Redundancy Effect (DRE), the extra F1 drop under repeated content relative to an approximately length-matched unrelated-filler control, and use it as a diagnostic contrast rather than a causal estimate. The effect is scale-dependent: DRE is large at 8B and near zero at 70B and Qwen2.5-72B, where F1 does not degrade. Validity metrics stay at 1.00 in both regimes, so they carry no information about content fidelity. A smaller synthetic news experiment shows the same dissociation with a different schema. We also give a zero-ground-truth drift check and monitoring recommendations. All evidence is synthetic, and we make no claim about real clinical notes.</abstract>
<identifier type="citekey">ram-etal-2026-validity</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.107/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>2112</start>
<end>2126</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T The Validity–Fidelity Gap: A Controlled Study of Representation Drift in Structured Extraction Under Redundant Contexts
%A Ram, Gokul Srinath Seetha
%A Vasudevan, Anirudh
%A Elavazhagan, Rashmi
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F ram-etal-2026-validity
%X Pipelines that use LLMs for structured extraction are commonly monitored by checking that each output parses as JSON and matches the expected keys. We show that this check can stay perfect while the extracted content degrades. We call the failure representation drift: the model still finds an entity but writes it in a different form than the intended one, for example a nested key-value object where a plain string is expected. The output is still valid JSON, yet the entity is scored both as missing (a false negative) and as an unexpected extra (a false positive). We study this with synthetic clinical notes repeated verbatim 1x-32x, which gives exact ground truth with prompt, schema and decoding held fixed. For LLaMA-3.1-8B-Instruct, micro-F1 falls from 0.499 to 0.259 while JSON validity stays 1.00. All 280 medication entities scored as missing in a 100-patient run are present in the output as nested objects, and a follow-up request with no schema recovers 96.7% of medications versus 40.0% under the schema at 16x. We introduce the Differential Redundancy Effect (DRE), the extra F1 drop under repeated content relative to an approximately length-matched unrelated-filler control, and use it as a diagnostic contrast rather than a causal estimate. The effect is scale-dependent: DRE is large at 8B and near zero at 70B and Qwen2.5-72B, where F1 does not degrade. Validity metrics stay at 1.00 in both regimes, so they carry no information about content fidelity. A smaller synthetic news experiment shows the same dissociation with a different schema. We also give a zero-ground-truth drift check and monitoring recommendations. All evidence is synthetic, and we make no claim about real clinical notes.
%U https://aclanthology.org/2026.findings-aacl.107/
%P 2112-2126
Markdown (Informal)
[The Validity–Fidelity Gap: A Controlled Study of Representation Drift in Structured Extraction Under Redundant Contexts](https://aclanthology.org/2026.findings-aacl.107/) (Ram et al., Findings 2026)
ACL