Saved in:
Bibliographic Details
Main Authors: Arviv, Dor, Elisha, Yehonatan, Barkan, Oren, Koenigstein, Noam
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.18024
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866908669685268480
author Arviv, Dor
Elisha, Yehonatan
Barkan, Oren
Koenigstein, Noam
author_facet Arviv, Dor
Elisha, Yehonatan
Barkan, Oren
Koenigstein, Noam
contents We present a method for extracting \emph{monosemantic} neurons, defined as latent dimensions that align with coherent and interpretable concepts, from user and item embeddings in recommender systems. Our approach employs a Sparse Autoencoder (SAE) to reveal semantic structure within pretrained representations. In contrast to work on language models, monosemanticity in recommendation must preserve the interactions between separate user and item embeddings. To achieve this, we introduce a \emph{prediction aware} training objective that backpropagates through a frozen recommender and aligns the learned latent structure with the model's user-item affinity predictions. The resulting neurons capture properties such as genre, popularity, and temporal trends, and support post hoc control operations including targeted filtering and content promotion without modifying the base model. Our method generalizes across different recommendation models and datasets, providing a practical tool for interpretable and controllable personalization. Code and evaluation resources are available at https://github.com/DeltaLabTLV/Monosemanticity4Rec.
format Preprint
id arxiv_https___arxiv_org_abs_2511_18024
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Extracting Interaction-Aware Monosemantic Concepts in Recommender Systems
Arviv, Dor
Elisha, Yehonatan
Barkan, Oren
Koenigstein, Noam
Information Retrieval
Artificial Intelligence
Machine Learning
We present a method for extracting \emph{monosemantic} neurons, defined as latent dimensions that align with coherent and interpretable concepts, from user and item embeddings in recommender systems. Our approach employs a Sparse Autoencoder (SAE) to reveal semantic structure within pretrained representations. In contrast to work on language models, monosemanticity in recommendation must preserve the interactions between separate user and item embeddings. To achieve this, we introduce a \emph{prediction aware} training objective that backpropagates through a frozen recommender and aligns the learned latent structure with the model's user-item affinity predictions. The resulting neurons capture properties such as genre, popularity, and temporal trends, and support post hoc control operations including targeted filtering and content promotion without modifying the base model. Our method generalizes across different recommendation models and datasets, providing a practical tool for interpretable and controllable personalization. Code and evaluation resources are available at https://github.com/DeltaLabTLV/Monosemanticity4Rec.
title Extracting Interaction-Aware Monosemantic Concepts in Recommender Systems
topic Information Retrieval
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2511.18024