Applying large-scale pre-trained visual models like CLIP to few-shot action recognition tasks can benefit performance and efficiency. Utilizing the "pre-training, fine-tuning" paradigm makes it possible to avoid training a network from scratch, which can be time-consuming and resource-intensive. However, this method has two drawbacks. First, limited labeled samples for few-shot action recognition necessitate minimizing the number of tunable parameters to mitigate over-fitting, also leading to inadequate fine-tuning that increases resource consumption and may disrupt the generalized representation of models. Second, the video's extra-temporal dimension challenges few-shot recognition's effective temporal modeling, while pre-trained visual models are usually image models. This paper proposes a novel method called Multimodal Adaptation of CLIP (MA-CLIP) to address these issues. It adapts CLIP for few-shot action recognition by adding lightweight adapters, which can minimize the number of learnable parameters and enable the model to transfer across different tasks quickly. The adapters we design can combine information from video-text multimodal sources for task-oriented spatiotemporal modeling, which is fast, efficient, and has low training costs. Additionally, based on the attention mechanism, we design a text-guided prototype construction module that can fully utilize video-text information to enhance the representation of video prototypes. Our MA-CLIP is plug-and-play, which can be used in any different few-shot action recognition temporal alignment metric.
Language Models today provide a high accuracy across a large number of downstream tasks. However, they remain susceptible to adversarial attacks, particularly against those where the adversarial examples maintain considerable similarity to the original text. Given the multilingual nature of text, the effectiveness of adversarial examples across translations and how machine translations can improve the robustness of adversarial examples remain largely unexplored. In this paper, we present a comprehensive study on the robustness of current text adversarial attacks to round-trip translation. We demonstrate that 6 state-of-the-art text-based adversarial attacks do not maintain their efficacy after round-trip translation. Furthermore, we introduce an intervention-based solution to this problem, by integrating Machine Translation into the process of adversarial example generation and demonstrating increased robustness to round-trip translation. Our results indicate that finding adversarial examples robust to translation can help identify the insufficiency of language models that is common across languages, and motivate further research into multilingual adversarial attacks.
Recent advances in visual-language models have shown remarkable zero-shot text-image matching ability that is transferable to down-stream tasks such as object detection and segmentation. However, adapting these models for object counting, which involves estimating the number of objects in an image, remains a formidable challenge. In this study, we conduct the first exploration of transferring visual-language models for class-agnostic object counting. Specifically, we propose CLIP-Count, a novel pipeline that estimates density maps for open-vocabulary objects with text guidance in a zero-shot manner, without requiring any finetuning on specific object classes. To align the text embedding with dense image features, we introduce a patch-text contrastive loss that guides the model to learn informative patch-level image representations for dense prediction. Moreover, we design a hierarchical patch-text interaction module that propagates semantic information across different resolution levels of image features. Benefiting from the full exploitation of the rich image-text alignment knowledge of pretrained visual-language models, our method effectively generates high-quality density maps for objects-of-interest. Extensive experiments on FSC-147, CARPK, and ShanghaiTech crowd counting datasets demonstrate that our proposed method achieves state-of-the-art accuracy and generalizability for zero-shot object counting. Project page at https://github.com/songrise/CLIP-Count
Image-text contrastive models such as CLIP are useful for a variety of downstream applications including zero-shot classification, image-text retrieval and transfer learning. However, these contrastively trained vision-language models often fail on compositional visio-linguistic tasks such as Winoground with performance equivalent to random chance. In our paper, we address this issue and propose a sample-efficient light-weight method called SDS-CLIP to improve the compositional visio-linguistic reasoning capabilities of CLIP. The core idea of our method is to use differentiable image parameterizations to fine-tune CLIP with a distillation objective from large text-to-image generative models such as Stable-Diffusion which are relatively good at visio-linguistic reasoning tasks. On the challenging Winoground compositional reasoning benchmark, our method improves the absolute visio-linguistic performance of different CLIP models by up to 7%, while on the ARO dataset, our method improves the visio-linguistic performance by upto 3%. As a byproduct of inducing visio-linguistic reasoning into CLIP, we also find that the zero-shot performance improves marginally on a variety of downstream datasets. Our method reinforces that carefully designed distillation objectives from generative models can be leveraged to extend existing contrastive image-text models with improved visio-linguistic reasoning capabilities.
Data augmentation is widely used in text classification, especially in the low-resource regime where a few examples for each class are available during training. Despite the success, generating data augmentations as hard positive examples that may increase their effectiveness is under-explored. This paper proposes an Adversarial Word Dilution (AWD) method that can generate hard positive examples as text data augmentations to train the low-resource text classification model efficiently. Our idea of augmenting the text data is to dilute the embedding of strong positive words by weighted mixing with unknown-word embedding, making the augmented inputs hard to be recognized as positive by the classification model. We adversarially learn the dilution weights through a constrained min-max optimization process with the guidance of the labels. Empirical studies on three benchmark datasets show that AWD can generate more effective data augmentations and outperform the state-of-the-art text data augmentation methods. The additional analysis demonstrates that the data augmentations generated by AWD are interpretable and can flexibly extend to new examples without further training.
Text classification in education, usually called auto-tagging, is the automated process of assigning relevant tags to educational content, such as questions and textbooks. However, auto-tagging suffers from a data scarcity problem, which stems from two major challenges: 1) it possesses a large tag space and 2) it is multi-label. Though a retrieval approach is reportedly good at low-resource scenarios, there have been fewer efforts to directly address the data scarcity problem. To mitigate these issues, here we propose a novel retrieval approach CEAA that provides effective learning in educational text classification. Our main contributions are as follows: 1) we leverage transfer learning from question-answering datasets, and 2) we propose a simple but effective data augmentation method introducing cross-encoder style texts to a bi-encoder architecture for more efficient inference. An extensive set of experiments shows that our proposed method is effective in multi-label scenarios and low-resource tags compared to state-of-the-art models.
Recently, speech-text pre-training methods have shown remarkable success in many speech and natural language processing tasks. However, most previous pre-trained models are usually tailored for one or two specific tasks, but fail to conquer a wide range of speech-text tasks. In addition, existing speech-text pre-training methods fail to explore the contextual information within a dialogue to enrich utterance representations. In this paper, we propose Speech-text dialog Pre-training for spoken dialog understanding with ExpliCiT cRoss-Modal Alignment (SPECTRA), which is the first-ever speech-text dialog pre-training model. Concretely, to consider the temporality of speech modality, we design a novel temporal position prediction task to capture the speech-text alignment. This pre-training task aims to predict the start and end time of each textual word in the corresponding speech waveform. In addition, to learn the characteristics of spoken dialogs, we generalize a response selection task from textual dialog pre-training to speech-text dialog pre-training scenarios. Experimental results on four different downstream speech-text tasks demonstrate the superiority of SPECTRA in learning speech-text alignment and multi-turn dialog context.
Natural language processing (NLP) models have become increasingly popular in real-world applications, such as text classification. However, they are vulnerable to privacy attacks, including data reconstruction attacks that aim to extract the data used to train the model. Most previous studies on data reconstruction attacks have focused on LLM, while classification models were assumed to be more secure. In this work, we propose a new targeted data reconstruction attack called the Mix And Match attack, which takes advantage of the fact that most classification models are based on LLM. The Mix And Match attack uses the base model of the target model to generate candidate tokens and then prunes them using the classification head. We extensively demonstrate the effectiveness of the attack using both random and organic canaries. This work highlights the importance of considering the privacy risks associated with data reconstruction attacks in classification models and offers insights into possible leakages.
We observe that the mapping between an image's representation in one model to its representation in another can be learned surprisingly well with just a linear layer, even across diverse models. Building on this observation, we propose $\textit{text-to-concept}$, where features from a fixed pretrained model are aligned linearly to the CLIP space, so that text embeddings from CLIP's text encoder become directly comparable to the aligned features. With text-to-concept, we convert fixed off-the-shelf vision encoders to surprisingly strong zero-shot classifiers for free, with accuracy at times even surpassing that of CLIP, despite being much smaller models and trained on a small fraction of the data compared to CLIP. We show other immediate use-cases of text-to-concept, like building concept bottleneck models with no concept supervision, diagnosing distribution shifts in terms of human concepts, and retrieving images satisfying a set of text-based constraints. Lastly, we demonstrate the feasibility of $\textit{concept-to-text}$, where vectors in a model's feature space are decoded by first aligning to the CLIP before being fed to a GPT-based generative model. Our work suggests existing deep models, with presumably diverse architectures and training, represent input samples relatively similarly, and a two-way communication across model representation spaces and to humans (through language) is viable.
The remarkable multimodal capabilities demonstrated by OpenAI's GPT-4 have sparked significant interest in the development of multimodal Large Language Models (LLMs). A primary research objective of such models is to align visual and textual modalities effectively while comprehending human instructions. Current methodologies often rely on annotations derived from benchmark datasets to construct image-dialogue datasets for training purposes, akin to instruction tuning in LLMs. However, these datasets often exhibit domain bias, potentially constraining the generative capabilities of the models. In an effort to mitigate these limitations, we propose a novel data collection methodology that synchronously synthesizes images and dialogues for visual instruction tuning. This approach harnesses the power of generative models, marrying the abilities of ChatGPT and text-to-image generative models to yield a diverse and controllable dataset with varied image content. This not only provides greater flexibility compared to existing methodologies but also significantly enhances several model capabilities. Our research includes comprehensive experiments conducted on various datasets using the open-source LLAVA model as a testbed for our proposed pipeline. Our results underscore marked enhancements across more than ten commonly assessed capabilities,