HPT++: Hierarchically Prompting Vision-Language Models with Multi-Granularity Knowledge Generation and Improved Structure Modeling
This work improves prompt tuning for vision-language models, which is important for researchers and practitioners in multimodal AI, though it appears incremental as it builds on prior prompt learning approaches.
The paper tackles the problem of adapting vision-language models to downstream tasks by addressing the lack of structured knowledge in conventional prompt descriptions, proposing HPT++ which improves performance across various evaluation settings and consistently outperforms existing state-of-the-art methods.
Prompt learning has become a prevalent strategy for adapting vision-language foundation models (VLMs) such as CLIP to downstream tasks. With the emergence of large language models (LLMs), recent studies have explored the potential of using category-related descriptions to enhance prompt effectiveness. However, conventional descriptions lack explicit structured information necessary to represent the interconnections among key elements like entities or attributes with relation to a particular category. Since existing prompt tuning methods give little consideration to managing structured knowledge, this paper advocates leveraging LLMs to construct a graph for each description to prioritize such structured knowledge. Consequently, we propose a novel approach called Hierarchical Prompt Tuning (HPT), enabling simultaneous modeling of both structured and conventional linguistic knowledge. Specifically, we introduce a relationship-guided attention module to capture pair-wise associations among entities and attributes for low-level prompt learning. In addition, by incorporating high-level and global-level prompts modeling overall semantics, the proposed hierarchical structure forges cross-level interlinks and empowers the model to handle more complex and long-term relationships. Finally, by enhancing multi-granularity knowledge generation, redesigning the relationship-driven attention re-weighting module, and incorporating consistent constraints on the hierarchical text encoder, we propose HPT++, which further improves the performance of HPT. Our experiments are conducted across a wide range of evaluation settings, including base-to-new generalization, cross-dataset evaluation, and domain generalization. Extensive results and ablation studies demonstrate the effectiveness of our methods, which consistently outperform existing SOTA methods.