Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Liu, Yuejiang, Alahi, Alexandre
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914689873608704
author Liu, Yuejiang
Alahi, Alexandre
author_facet Liu, Yuejiang
Alahi, Alexandre
contents Steering the behavior of a strong model pre-trained on internet-scale data can be difficult due to the scarcity of competent supervisors. Recent studies reveal that, despite supervisory noises, a strong student model may surpass its weak teacher when fine-tuned on specific objectives. Yet, the effectiveness of such weak-to-strong generalization remains limited, especially in the presence of large capability gaps. In this paper, we propose to address this challenge by harnessing a diverse set of specialized teachers, instead of a single generalist one, that collectively supervises the strong student. Our approach resembles the classical hierarchical mixture of experts, with two components tailored for co-supervision: (i) we progressively alternate student training and teacher assignment, leveraging the growth of the strong student to identify plausible supervisions; (ii) we conservatively enforce teacher-student and local-global consistency, leveraging their dependencies to reject potential annotation noises. We validate the proposed method through visual recognition tasks on the OpenAI weak-to-strong benchmark and additional multi-domain datasets. Our code is available at \url{https://github.com/yuejiangliu/csl}.
format Preprint
id arxiv_https___arxiv_org_abs_2402_15505
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts
Liu, Yuejiang
Alahi, Alexandre
Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
Steering the behavior of a strong model pre-trained on internet-scale data can be difficult due to the scarcity of competent supervisors. Recent studies reveal that, despite supervisory noises, a strong student model may surpass its weak teacher when fine-tuned on specific objectives. Yet, the effectiveness of such weak-to-strong generalization remains limited, especially in the presence of large capability gaps. In this paper, we propose to address this challenge by harnessing a diverse set of specialized teachers, instead of a single generalist one, that collectively supervises the strong student. Our approach resembles the classical hierarchical mixture of experts, with two components tailored for co-supervision: (i) we progressively alternate student training and teacher assignment, leveraging the growth of the strong student to identify plausible supervisions; (ii) we conservatively enforce teacher-student and local-global consistency, leveraging their dependencies to reject potential annotation noises. We validate the proposed method through visual recognition tasks on the OpenAI weak-to-strong benchmark and additional multi-domain datasets. Our code is available at \url{https://github.com/yuejiangliu/csl}.
title Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts
topic Machine Learning
Artificial Intelligence
Computer Vision and Pattern Recognition
url https://arxiv.org/abs/2402.15505