@inproceedings{mercorio-etal-2025-sfal,
title = "{SFAL}: Semantic-Functional Alignment Scores for Distributional Evaluation of Auto-Interpretability in Sparse Autoencoders",
author = "Mercorio, Fabio and
Pallucchini, Filippo and
Potert{\`i}, Daniele and
Serino, Antonio and
Seveso, Andrea",
editor = "Potdar, Saloni and
Rojas-Barahona, Lina and
Montella, Sebastien",
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track",
month = nov,
year = "2025",
address = "Suzhou (China)",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.emnlp-industry.39/",
doi = "10.18653/v1/2025.emnlp-industry.39",
pages = "576--583",
ISBN = "979-8-89176-333-3",
abstract = "Interpreting the internal representations of large language models (LLMs) is crucial for their deployment in real-world applications, impacting areas such as AI safety, debugging, and compliance. Sparse Autoencoders facilitate interpretability by decomposing polysemantic activation into a latent space of monosemantic features. However, evaluating the auto-interpretability of these features is difficult and computationally expensive, which limits scalability in practical settings. In this work, we propose SFAL, an alternative evaluation strategy that reduces reliance on LLM-based scoring by assessing the alignment between the semantic neighbourhoods of features (derived from auto-interpretation embeddings) and their functional neighbourhoods (derived from co-occurrence statistics).Our method enhances efficiency, enabling fast and cost-effective assessments. We validate our approach on large-scale models, demonstrating its potential to provide interpretability while reducing computational overhead, making it suitable for real-world deployment."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="mercorio-etal-2025-sfal">
<titleInfo>
<title>SFAL: Semantic-Functional Alignment Scores for Distributional Evaluation of Auto-Interpretability in Sparse Autoencoders</title>
</titleInfo>
<name type="personal">
<namePart type="given">Fabio</namePart>
<namePart type="family">Mercorio</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Filippo</namePart>
<namePart type="family">Pallucchini</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Daniele</namePart>
<namePart type="family">Potertì</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Antonio</namePart>
<namePart type="family">Serino</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Andrea</namePart>
<namePart type="family">Seveso</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track</title>
</titleInfo>
<name type="personal">
<namePart type="given">Saloni</namePart>
<namePart type="family">Potdar</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Lina</namePart>
<namePart type="family">Rojas-Barahona</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sebastien</namePart>
<namePart type="family">Montella</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou (China)</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-333-3</identifier>
</relatedItem>
<abstract>Interpreting the internal representations of large language models (LLMs) is crucial for their deployment in real-world applications, impacting areas such as AI safety, debugging, and compliance. Sparse Autoencoders facilitate interpretability by decomposing polysemantic activation into a latent space of monosemantic features. However, evaluating the auto-interpretability of these features is difficult and computationally expensive, which limits scalability in practical settings. In this work, we propose SFAL, an alternative evaluation strategy that reduces reliance on LLM-based scoring by assessing the alignment between the semantic neighbourhoods of features (derived from auto-interpretation embeddings) and their functional neighbourhoods (derived from co-occurrence statistics).Our method enhances efficiency, enabling fast and cost-effective assessments. We validate our approach on large-scale models, demonstrating its potential to provide interpretability while reducing computational overhead, making it suitable for real-world deployment.</abstract>
<identifier type="citekey">mercorio-etal-2025-sfal</identifier>
<identifier type="doi">10.18653/v1/2025.emnlp-industry.39</identifier>
<location>
<url>https://aclanthology.org/2025.emnlp-industry.39/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>576</start>
<end>583</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T SFAL: Semantic-Functional Alignment Scores for Distributional Evaluation of Auto-Interpretability in Sparse Autoencoders
%A Mercorio, Fabio
%A Pallucchini, Filippo
%A Potertì, Daniele
%A Serino, Antonio
%A Seveso, Andrea
%Y Potdar, Saloni
%Y Rojas-Barahona, Lina
%Y Montella, Sebastien
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing: Industry Track
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou (China)
%@ 979-8-89176-333-3
%F mercorio-etal-2025-sfal
%X Interpreting the internal representations of large language models (LLMs) is crucial for their deployment in real-world applications, impacting areas such as AI safety, debugging, and compliance. Sparse Autoencoders facilitate interpretability by decomposing polysemantic activation into a latent space of monosemantic features. However, evaluating the auto-interpretability of these features is difficult and computationally expensive, which limits scalability in practical settings. In this work, we propose SFAL, an alternative evaluation strategy that reduces reliance on LLM-based scoring by assessing the alignment between the semantic neighbourhoods of features (derived from auto-interpretation embeddings) and their functional neighbourhoods (derived from co-occurrence statistics).Our method enhances efficiency, enabling fast and cost-effective assessments. We validate our approach on large-scale models, demonstrating its potential to provide interpretability while reducing computational overhead, making it suitable for real-world deployment.
%R 10.18653/v1/2025.emnlp-industry.39
%U https://aclanthology.org/2025.emnlp-industry.39/
%U https://doi.org/10.18653/v1/2025.emnlp-industry.39
%P 576-583
Markdown (Informal)
[SFAL: Semantic-Functional Alignment Scores for Distributional Evaluation of Auto-Interpretability in Sparse Autoencoders](https://aclanthology.org/2025.emnlp-industry.39/) (Mercorio et al., EMNLP 2025)
ACL