Fantastic Features and Where to Find Them: A Probing Method to combine Features from Multiple Foundation Models

TL;DR

This paper introduces ComBo, a scalable probing method that combines features from multiple foundation models without dataset-specific tuning, improving task performance and enabling efficient model comparison.

Contribution

ComBo is a novel, scalable probing-based adapter that integrates features from multiple foundation models and layers without backpropagation, enhancing multi-model representation and task-specific performance.

Findings

Outperforms previous probing methods on VTAB-1k tasks

Matches or surpasses distillation-based model merging techniques

Enables efficient evaluation of model relevance for different tasks

Abstract

Foundation models (FMs) trained with different objectives and data learn diverse representations, making some more effective than others for specific downstream tasks. Existing adaptation strategies, such as parameter-efficient fine-tuning, focus on individual models and do not exploit the complementary strengths across models. Probing methods offer a promising alternative by extracting information from frozen models, but current techniques do not scale well with large feature sets and often rely on dataset-specific hyperparameter tuning. We propose Combined backBones (ComBo), a simple and scalable probing-based adapter that effectively integrates features from multiple models and layers. ComBo compresses activations from layers of one or more FMs into compact token-wise representations and processes them with a lightweight transformer for task-specific prediction. Crucially, ComBo does not require dataset-specific tuning or backpropagation through the backbone models. However, not all models are equally relevant for all tasks. To address this, we introduce a mechanism that leverages ComBo's joint multi-backbone probing to efficiently evaluate each backbone's task-relevance, enabling both practical model comparison and improved performance through selective adaptation. On the 19 tasks of the VTAB-1k benchmark, ComBo outperforms previous probing methods, matches or surpasses more expensive alternatives, such as distillation-based model merging, and enables efficient probing of tuned models. Our results demonstrate that ComBo offers a practical and general-purpose framework for combining diverse representations from multiple FMs.