Learning Task-Agnostic Representations through Multi-Teacher Distillation

Fuente: arXiv
Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Formont, Philippe, Darrin, Maxime, Karimian, Banafsheh, Cheung, Jackie CK, Granger, Eric, Ayed, Ismail Ben, Shateri, Mohammadhadi, Piantanida, Pablo
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866915567719415808
author Formont, Philippe
Darrin, Maxime
Karimian, Banafsheh
Cheung, Jackie CK
Granger, Eric
Ayed, Ismail Ben
Shateri, Mohammadhadi
Piantanida, Pablo
author_facet Formont, Philippe
Darrin, Maxime
Karimian, Banafsheh
Cheung, Jackie CK
Granger, Eric
Ayed, Ismail Ben
Shateri, Mohammadhadi
Piantanida, Pablo
contents Casting complex inputs into tractable representations is a critical step across various fields. Diverse embedding models emerge from differences in architectures, loss functions, input modalities and datasets, each capturing unique aspects of the input. Multi-teacher distillation leverages this diversity to enrich representations but often remains tailored to specific tasks. In this paper, we introduce a task-agnostic framework based on a ``majority vote" objective function. We demonstrate that this function is bounded by the mutual information between student and teachers' embeddings, leading to a task-agnostic distillation loss that eliminates dependence on task-specific labels or prior knowledge. Our evaluations across text, vision models, and molecular modeling show that our method effectively leverages teacher diversity, resulting in representations enabling better performance for a wide range of downstream tasks such as classification, clustering, or regression. Additionally, we train and release state-of-the-art embedding models, enhancing downstream performance in various modalities.
format Preprint
id arxiv_https___arxiv_org_abs_2510_18680
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning Task-Agnostic Representations through Multi-Teacher Distillation
Formont, Philippe
Darrin, Maxime
Karimian, Banafsheh
Cheung, Jackie CK
Granger, Eric
Ayed, Ismail Ben
Shateri, Mohammadhadi
Piantanida, Pablo
Machine Learning
Casting complex inputs into tractable representations is a critical step across various fields. Diverse embedding models emerge from differences in architectures, loss functions, input modalities and datasets, each capturing unique aspects of the input. Multi-teacher distillation leverages this diversity to enrich representations but often remains tailored to specific tasks. In this paper, we introduce a task-agnostic framework based on a ``majority vote" objective function. We demonstrate that this function is bounded by the mutual information between student and teachers' embeddings, leading to a task-agnostic distillation loss that eliminates dependence on task-specific labels or prior knowledge. Our evaluations across text, vision models, and molecular modeling show that our method effectively leverages teacher diversity, resulting in representations enabling better performance for a wide range of downstream tasks such as classification, clustering, or regression. Additionally, we train and release state-of-the-art embedding models, enhancing downstream performance in various modalities.
title Learning Task-Agnostic Representations through Multi-Teacher Distillation
topic Machine Learning
url https://arxiv.org/abs/2510.18680