Multiple Stochastic Prompt Tuning for Practical Cross-Domain Few Shot Learning

In this work, we propose a practical cross-domain few-shot learning (pCDFSL) task, where a large-scale pre-trained model like CLIP can be easily deployed on a target dataset. The goal is to simultaneously classify all unseen classes under extreme domain shifts, by utilizing only a few labeled samples per class. The pCDFSL paradigm is source-free and moves beyond artificially created episodic training and testing regimes followed by existing CDFSL frameworks, making it more challenging and relevant to real-world applications. Towards that goal, we propose a novel framework, termed MIST (MultIple STochastic Prompt tuning), where multiple stochastic prompts are utilized to handle significant domain and semantic shifts. Specifically, multiple prompts are learnt for each class, effectively capturing multiple peaks in the input data. Furthermore, instead of representing the weights of the multiple prompts as point-estimates, we model them as learnable Gaussian distributions with two different strategies, encouraging an efficient exploration of the prompt parameter space, which mitigate overfitting due to the few labeled training samples. Extensive experiments and comparison with the state-of-the-art methods on four CDFSL benchmarks adapted to this setting, show the effectiveness of the proposed framework.

Prompt Tuning Vision Language Models with Margin Regularizer for Few-Shot Learning under Distribution Shifts

Recently, Vision-Language foundation models like CLIP and ALIGN, which are pre-trained on large-scale data have shown remarkable zero-shot generalization to diverse datasets with different classes and even domains. In this work, we take a step further and analyze whether these models can be adapted to target datasets having very different distributions and classes compared to what these models have been trained on, using only a few labeled examples from the target dataset. In such scenarios, finetuning large pretrained models is challenging due to problems of overfitting as well as loss of generalization, and has not been well explored in prior literature. Since, the pre-training data of such models are unavailable, it is difficult to comprehend the performance on various downstream datasets. First, we try to answer the question: Given a target dataset with a few labelled examples, can we estimate whether further fine-tuning can enhance the performance compared to zero-shot evaluation? by analyzing the common vision-language embedding space. Based on the analysis, we propose a novel prompt-tuning method, PromptMargin for adapting such large-scale VLMs directly on the few target samples. PromptMargin effectively tunes the text as well as visual prompts for this task, and has two main modules: 1) Firstly, we use a selective augmentation strategy to complement the few training samples in each task; 2) Additionally, to ensure robust training in the presence of unfamiliar class names, we increase the inter-class margin for improved class discrimination using a novel Multimodal Margin Regularizer. Extensive experiments and analysis across fifteen target benchmark datasets, with varying degrees of distribution shifts from natural images, shows the effectiveness of the proposed framework over the existing state-of-the-art approaches applied to this setting. github.com/debarshigit/PromptMargin.

Leveraging Out-of-Domain Data for Domain-Specific Prompt Tuning in Multi-Modal Fake News Detection

The spread of fake news using out-of-context images has become widespread and is a challenging task in this era of information overload. Since annotating huge amounts of such data requires significant time of domain experts, it is imperative to develop methods which can work in limited annotated data scenarios. In this work, we explore whether out-of-domain data can help to improve out-of-context misinformation detection (termed here as multi-modal fake news detection) of a desired domain, eg. politics, healthcare, etc. Towards this goal, we propose a novel framework termed DPOD (Domain-specific Prompt-tuning using Out-of-Domain data). First, to compute generalizable features, we modify the Vision-Language Model, CLIP to extract features that helps to align the representations of the images and corresponding text captions of both the in-domain and out-of-domain data in a label-aware manner. Further, we propose a domain-specific prompt learning technique which leverages the training samples of all the available domains based on the the extent they can be useful to the desired domain. Extensive experiments on a large-scale benchmark dataset, namely NewsClippings demonstrate that the proposed framework achieves state of-the-art performance, significantly surpassing the existing approaches for this challenging task.