@inproceedings{casola-etal-2026-steereval,
title = "{STEEREVAL}: Inference-time Interventions Strengthen Multilingual Generalization in Neural Summarization Metrics",
author = {Casola, Silvia and
Shim, Ryan Soh-Eun and
K{\"o}rner, Felicia and
Mao, Yuchen and
Plank, Barbara},
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.2486/",
pages = "56439--56466",
ISBN = "979-8-89176-459-0",
abstract = "An empirical bottleneck for Natural Language Generation tasks such as summarization is the lack of accurate and robust evaluation metrics for many languages, which hinders progress. While Large Language Models (LLMs) are increasingly adopted as evaluators with the advent of LLM-as-a-Judge approaches, their agreement with human judgments remains low across many languages. We take inspiration from recent studies suggesting that multilingual language models often use English as an internal pivot language, and that misalignment with this pivot can lead to degraded downstream performance during text generation. Motivated by this hypothesis, we ask whether steering the activations of neural metrics with respect to an English pivot can improve their correlation with human judgments in multilingual summarization evaluation. We experiment with encoder- and decoder-based metrics across five languages. We find that activation steering is effective for neural metrics, and that a single, language-independent steering vector improves metric{--}human agreement. Steering also results in improvements when applied to COMET, a small encoder-based metric, showing notable human agreement despite its small scale."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="casola-etal-2026-steereval">
<titleInfo>
<title>STEEREVAL: Inference-time Interventions Strengthen Multilingual Generalization in Neural Summarization Metrics</title>
</titleInfo>
<name type="personal">
<namePart type="given">Silvia</namePart>
<namePart type="family">Casola</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ryan</namePart>
<namePart type="given">Soh-Eun</namePart>
<namePart type="family">Shim</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Felicia</namePart>
<namePart type="family">Körner</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuchen</namePart>
<namePart type="family">Mao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Barbara</namePart>
<namePart type="family">Plank</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>An empirical bottleneck for Natural Language Generation tasks such as summarization is the lack of accurate and robust evaluation metrics for many languages, which hinders progress. While Large Language Models (LLMs) are increasingly adopted as evaluators with the advent of LLM-as-a-Judge approaches, their agreement with human judgments remains low across many languages. We take inspiration from recent studies suggesting that multilingual language models often use English as an internal pivot language, and that misalignment with this pivot can lead to degraded downstream performance during text generation. Motivated by this hypothesis, we ask whether steering the activations of neural metrics with respect to an English pivot can improve their correlation with human judgments in multilingual summarization evaluation. We experiment with encoder- and decoder-based metrics across five languages. We find that activation steering is effective for neural metrics, and that a single, language-independent steering vector improves metric–human agreement. Steering also results in improvements when applied to COMET, a small encoder-based metric, showing notable human agreement despite its small scale.</abstract>
<identifier type="citekey">casola-etal-2026-steereval</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2486/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>56439</start>
<end>56466</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T STEEREVAL: Inference-time Interventions Strengthen Multilingual Generalization in Neural Summarization Metrics
%A Casola, Silvia
%A Shim, Ryan Soh-Eun
%A Körner, Felicia
%A Mao, Yuchen
%A Plank, Barbara
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F casola-etal-2026-steereval
%X An empirical bottleneck for Natural Language Generation tasks such as summarization is the lack of accurate and robust evaluation metrics for many languages, which hinders progress. While Large Language Models (LLMs) are increasingly adopted as evaluators with the advent of LLM-as-a-Judge approaches, their agreement with human judgments remains low across many languages. We take inspiration from recent studies suggesting that multilingual language models often use English as an internal pivot language, and that misalignment with this pivot can lead to degraded downstream performance during text generation. Motivated by this hypothesis, we ask whether steering the activations of neural metrics with respect to an English pivot can improve their correlation with human judgments in multilingual summarization evaluation. We experiment with encoder- and decoder-based metrics across five languages. We find that activation steering is effective for neural metrics, and that a single, language-independent steering vector improves metric–human agreement. Steering also results in improvements when applied to COMET, a small encoder-based metric, showing notable human agreement despite its small scale.
%U https://aclanthology.org/2026.emnlp-main.2486/
%P 56439-56466
Markdown (Informal)
[STEEREVAL: Inference-time Interventions Strengthen Multilingual Generalization in Neural Summarization Metrics](https://aclanthology.org/2026.emnlp-main.2486/) (Casola et al., EMNLP 2026)
ACL