Co-Supervised Learning: Improving Weak-to-Strong Generalization with Hierarchical Mixture of Experts
This work addresses a key challenge in AI alignment and model training, enabling more efficient use of weak supervision for strong models, though it is incremental as it builds on existing weak-to-strong generalization methods.
The paper tackles the problem of weak-to-strong generalization in AI, where strong models struggle to learn effectively from weak supervisors, by proposing a method that uses multiple specialized teachers to co-supervise the student, achieving improved performance on visual recognition tasks as validated on benchmarks like OpenAI's weak-to-strong dataset.
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}.