@inproceedings{farahani-etal-2026-copy,
title = "To Copy or Not to Copy: Copying Is Easier to Induce Than Recall",
author = "Farahani, Mehrdad and
Penzkofer, Franziska and
Johansson, Richard",
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.2329/",
doi = "10.18653/2026.findings-emnlp.2329",
pages = "50027--50090",
ISBN = "979-8-89176-462-0",
abstract = "Language models used in retrieval-augmented settings choose between parametric knowledge stored in their weights and contextual information in the prompt. This work presents a mechanistic study of that choice: using an approach based on steering vectors, we carry out interventions at selected layers and token spans, and we are able to toggle the model{'}s behavior in two directions: Copy{\textrightarrow}Recall (suppressing context use) and Recall{\textrightarrow}Copy (inducing the model to copy any token from the context). Experiments on three architectures (decoder-only and encoder/decoder) and two open-domain QA benchmarks show consistent behavior shifts under moderate scaling while monitoring accuracy and fluency. Mechanistic analyses of attention routing, MLP contributions, and layer-wise probability trajectories reveal an asymmetry: inducing copying is an easy ``reactivation'' process that can be triggered at different locations in the input, while restoring recall is a ``suppression'' process that is more fragile and strongly tied to object-token interventions."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="farahani-etal-2026-copy">
<titleInfo>
<title>To Copy or Not to Copy: Copying Is Easier to Induce Than Recall</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mehrdad</namePart>
<namePart type="family">Farahani</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Franziska</namePart>
<namePart type="family">Penzkofer</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Richard</namePart>
<namePart type="family">Johansson</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>Language models used in retrieval-augmented settings choose between parametric knowledge stored in their weights and contextual information in the prompt. This work presents a mechanistic study of that choice: using an approach based on steering vectors, we carry out interventions at selected layers and token spans, and we are able to toggle the model’s behavior in two directions: Copy→Recall (suppressing context use) and Recall→Copy (inducing the model to copy any token from the context). Experiments on three architectures (decoder-only and encoder/decoder) and two open-domain QA benchmarks show consistent behavior shifts under moderate scaling while monitoring accuracy and fluency. Mechanistic analyses of attention routing, MLP contributions, and layer-wise probability trajectories reveal an asymmetry: inducing copying is an easy “reactivation” process that can be triggered at different locations in the input, while restoring recall is a “suppression” process that is more fragile and strongly tied to object-token interventions.</abstract>
<identifier type="citekey">farahani-etal-2026-copy</identifier>
<identifier type="doi">10.18653/2026.findings-emnlp.2329</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.2329/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>50027</start>
<end>50090</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T To Copy or Not to Copy: Copying Is Easier to Induce Than Recall
%A Farahani, Mehrdad
%A Penzkofer, Franziska
%A Johansson, Richard
%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 farahani-etal-2026-copy
%X Language models used in retrieval-augmented settings choose between parametric knowledge stored in their weights and contextual information in the prompt. This work presents a mechanistic study of that choice: using an approach based on steering vectors, we carry out interventions at selected layers and token spans, and we are able to toggle the model’s behavior in two directions: Copy→Recall (suppressing context use) and Recall→Copy (inducing the model to copy any token from the context). Experiments on three architectures (decoder-only and encoder/decoder) and two open-domain QA benchmarks show consistent behavior shifts under moderate scaling while monitoring accuracy and fluency. Mechanistic analyses of attention routing, MLP contributions, and layer-wise probability trajectories reveal an asymmetry: inducing copying is an easy “reactivation” process that can be triggered at different locations in the input, while restoring recall is a “suppression” process that is more fragile and strongly tied to object-token interventions.
%R 10.18653/2026.findings-emnlp.2329
%U https://aclanthology.org/2026.findings-emnlp.2329/
%U https://doi.org/10.18653/2026.findings-emnlp.2329
%P 50027-50090
Markdown (Informal)
[To Copy or Not to Copy: Copying Is Easier to Induce Than Recall](https://aclanthology.org/2026.findings-emnlp.2329/) (Farahani et al., Findings 2026)
ACL