论文检索

Xianwei Zhuang, Xuxin Cheng, Zhihong Zhu 0001, Zhanpeng Chen, Hongxiang Li 0004, Yuexian Zou

Pre-trained language models (PLMs) that rely solely on textual corpus may present limitations in multimodal semantics comprehension. Existing studies attempt to alleviate this issue by incorporating additional modal information through image retrieval or generation. However, these methods: (1) inevitably encounter modality gaps and noise; (2) treat all modalities indiscriminately; and (3) ignore visual or acoustic semantics of key entities. To tackle these challenges, we propose a novel principled iterative framework for multimodal-augmented PLMs termed MASE, which achieves efficient and balanced injection of multimodal semantics under the proposed Expectation Maximization (EM) based iterative algorithm. Initially, MASE utilizes multimodal proxies instead of explicit data to enhance PLMs, which avoids noise and modality gaps. In E-step, MASE adopts a novel information-driven self-balanced strategy to estimate allocation weights. Furthermore, MASE employs heterogeneous graph attention to capture entity-level fine-grained semantics on the proposed multimodal-semantic scene graph. In M-step, MASE injects global multimodal knowledge into PLMs through a cross-modal contrastive loss. Experimental results show that MASE consistently outperforms competitive baselines on multiple tasks across various architectures. More impressively, MASE is compatible with existing efficient parameter fine-tuning methods, such as prompt learning.

Muquan Li, Dongyang Zhang 0001, Tao He 0007, Xiurui Xie, Yuan-Fang Li, Ke Qin

Data-free knowledge distillation (DFKD) has emerged as a pivotal technique in the domain of model compression, substantially reducing the dependency on the original training data. Nonetheless, conventional DFKD methods that employ synthesized training data are prone to the limitations of inadequate diversity and discrepancies in distribution between the synthesized and original datasets. To address these challenges, this paper introduces an innovative approach to DFKD through diverse diffusion augmentation (DDA). Specifically, we revise the paradigm of common data synthesis in DFKD to a composite process through leveraging diffusion models subsequent to data synthesis for self-supervised augmentation, which generates a spectrum of data samples with similar distributions while retaining controlled variations. Furthermore, to mitigate excessive deviation in the embedding space, we introduce an image filtering technique grounded in cosine similarity to maintain fidelity during the knowledge distillation process. Comprehensive experiments conducted on CIFAR-10, CIFAR-100, and Tiny-ImageNet datasets showcase the superior performance of our method across various teacher-student network configurations, outperforming the contemporary state-of-the-art DFKD methods. Code will be available at: https://github.com/SLGSP/DDA.

Dian Xie, Peiang Zhao, Jiarui Zhang, Kangqi Wei, Xiaobao Ni, Jiong Xia

Reconstructing visual stimuli from brain activities is crucial for deciphering the underlying mechanism of the human visual system. While recent studies have achieved notable results by leveraging deep generative models, challenges persist due to the lack of large-scale datasets and the inherent noise from non-invasive measurement methods. In this study, we draw inspiration from the mechanism of human memory and propose BrainRAM, a novel two-stage dual-guided framework for visual stimuli reconstruction. BrainRAM incorporates a Retrieval-Augmented Module (RAM) and diffusion prior to enhance the quality of reconstructed images from the brain. Specifically, in stage I, we transform fMRI voxels into the latent space of image and text embeddings via diffusion priors, obtaining preliminary estimates of the visual stimuli's semantics and structure. In stage II, based on previous estimates, we retrieve data from the LAION-2B-en dataset and employ the proposed RAM to refine them, yielding high-quality reconstruction results. Extensive experiments demonstrate that our BrainRAM outperforms current state-of-the-art methods both qualitatively and quantitatively, providing a new perspective for visual stimuli reconstruction.

Jingxiong Li, Sunyi Zheng, Chenglu Zhu, Yuxuan Sun 0002, Pingyi Chen, Zhongyi Shui, Yunlong Zhang, Honglin Li 0001, Lin Yang 0002

In digital pathology, cancer lesions are identified by analyzing the spatial context within pathology images. Synthesizing such complex spatial context is challenging as pathology whole slide images typically exhibit high resolution, low inter-class variety, and are sparsely labeled. To address these challenges, we propose PathUp, a novel diffusion model tailored for the synthesis of multi-class high-resolution pathology images. Our approach includes a latent space patch-wise timestep tracking, which helps to generate high-quality images without tiling artifacts. Pathology knowledge is integrated through our patho-align. The robust generation of lesion subtypes and scale information is ensured by introducing a feature entropy loss. The effectiveness of our method is evaluated through extensive experiments, supplemented by assessments from human experts, demonstrating the authenticity of the synthetic data produced. Furthermore, we highlight the potential utility of our generated images as an augmentation method, thereby enhancing the performance of downstream tasks such as cancer subtype classification.

Dongjie Fu, Xize Cheng, Xiaoda Yang, Hanting Wang, Zhou Zhao 0001, Tao Jin 0004

In the burgeoning field of Audio-Visual Speech Recognition (AVSR), extant research has predominantly concentrated on the training paradigms tailored for high-quality resources. However, owing to the challenges inherent in real-world data collection, audio-visual data are frequently affected by modality-distortion, which encompasses audio-visual asynchrony, video noise and audio noise. The recognition accuracy of existing AVSR method is significantly compromised when multiple modality-distortion coexist in low-resource data. In light of the above challenges, we propose PCD: cluster-Prompt with Contrastive Decomposition, a robust framework for modality-distortion speech recognition, specifically devised to transpose the pre-trained knowledge from high-resource domain to the targeted domain by leveraging contrast-augmented prompts. In contrast to previous studies, we take into consideration the possibility of various types of distortion in both the audio and visual modalities. Concretely, we design bespoke prompts to delineate each modality-distortion, guiding the model to achieve speech recognition applicable to various distortion scenarios with quite few learnable parameters. To materialize the prompt mechanism, we employ multiple cluster-based strategies that better suits the pre-trained audio-visual model. Additionally, we design a contrastive decomposition mechanism to restrict the explicit relationships among various modality conditions, given their shared task knowledge and disparate modality priors. Extensive results on LRS2 dataset demonstrate that PCD achieves state-of-the-art performance for audio-visual speech recognition under the constraints of distorted resources. Code is available at https://github.com/ballooncatt/PCD.

Haoyang Su, Wenzhe Du, Xiaoliang Wang 0001, Cam-Tu Nguyen

With the increasing prevalence of virtual assistants, multimodal conversational recommendation systems (multimodal CRS) becomes essential for boosting customer engagement, improving conversion rates, and enhancing user satisfaction. Yet conversational samples, as training data for such a system, are difficult to obtain in large quantities, particularly in new platforms. To effectively train multimodal CRS in a small data setting, we enhance data quality to make up for the small data quantity by augmenting conversations with dialogue states. We then devise an effective dialogue state encoder to bridge the semantic gap between conversation and product representations for recommendation. To further reduce the cost of dialogue state annotation, a semi-supervised learning method is developed to effectively train the dialogue state encoder with a small set of labeled conversations. In addition, we design a correlation regularisation that leverages knowledge in the multimodal product database to help align textual and visual modalities. Experiments on the dataset MMD demonstrate the effectiveness of our method. Particularly, with only 5% of the MMD training set, our method (namely SeMANTIC) obtains better NDCG scores than those of baseline models trained on the full MMD training set.

Wenlin Li, Yucheng Xu, Xiaoqing Zheng, Suoya Han, Jun Wang 0018, Xiaobo Sun

Sparse and noisy images (SNIs), like those in spatial gene expression data, pose significant challenges for effective representation learning and clustering, which are essential for thorough data analysis and interpretation. In response to these challenges, we propose Dual Advancement of Representation Learning and Clustering (DARLC), an innovative framework that leverages contrastive learning to enhance the representations derived from masked image modeling. Simultaneously, DARLC integrates cluster assignments in a cohesive, end-to-end approach. This integrated clustering strategy addresses the "class collision problem" inherent in contrastive learning, thus improving the quality of the resulting representations. To generate more plausible positive views for contrastive learning, we employ a graph attention network-based technique that produces denoised images as augmented data. As such, our framework offers a comprehensive approach that improves the learning of representations by enhancing their local perceptibility, distinctiveness, and the understanding of relational semantics. Furthermore, we utilize a Student's t mixture model to achieve more robust and adaptable clustering of SNIs. Extensive experiments, conducted across 12 different types of datasets consisting of SNIs, demonstrate that DARLC surpasses the state-of-the-art methods in both image clustering and generating image representations that accurately capture gene interactions. Code is available at https://github.com/zipging/DARLC.

Kang Xia, Wenzhong Li, Yimiao Shao, Sanglu Lu

Human Activity Recognition (HAR) as an emerging research field has attracted widespread academic attention due to its wide range of practical applications in areas such as healthcare, environmental monitoring, and sports training. Given the high cost of annotating sensor data, many unsupervised and semi-supervised methods have been applied to HAR to alleviate the problem of limited data. In this paper, we propose a novel video-enhanced cross-modal collaborative learning method, Vi2ACT, to address the issue of few-shot HAR. We introduce a new data augmentation approach that utilizes a text-to-video generation model to generate class-related videos. Subsequently, a large quantity of video semantic representations are obtained through fine-tuning the video encoder for cross-modal co-learning. Furthermore, to effectively align video semantic representations and time series representations, we enhance HAR at the representation-level using conditional Generative Adversarial Nets (cGAN). We design a novel Representation Conditional Discriminator that is trained to assess samples as originating from video representations rather than those generated by the time series encoder as accurately as possible. We conduct extensive experiments on four commonly used HAR datasets. The experimental results demonstrate that our method outperforms other baseline models in all few-shot scenarios.

Liang He 0009, Hongke Wang, Zhen Wu 0002, Jianbing Zhang, Xinyu Dai, Jiajun Chen 0001

Multimedia content's surge on the internet has made multimodal relation extraction vital for applications like intelligent search and knowledge graph construction. As a rich source of image-text data, social media plays a crucial role in populating knowledge bases. However, the noisy information present in social media poses a challenge in multimodal relation extraction. Current methods focus on extracting relevant information from images to improve model performance but often overlook the importance of global image information. In this paper, we propose a novel multimodal relation extraction method FocalMRE, which leverages image focal augmentation, focal attention, and gating mechanisms. FocalMRE enables the model to concentrate on the image's focal regions while effectively utilizing the global information in the image. Through gating mechanisms, FocalMRE optimizes the multimodal fusion strategy, allowing the model to select the most relevant augmented regions for overcoming noise interference in relation extraction. The experimental results on the public MNRE dataset reveal that FocalMRE exhibits robust and significant performance advantages in the multimodal relation extraction task, especially in scenarios with high noise, long-tail distributions, and limited resources. The code is available at https://github.com/NJUNLP/FocalMRE.

Minghang Zheng, Jiahua Zhang, Qingchao Chen, Yuxin Peng 0001, Yang Liu 0105

Visual grounding aims to localize the object referred to in an image based on a natural language query. Although progress has been made recently, accurately localizing target objects within multiple-instance distractions (multiple objects of the same category as the target) remains a significant challenge. Existing methods demonstrate a significant performance drop when there are multiple distractions in an image, indicating an insufficient understanding of the fine-grained semantics and spatial relationships between objects. In this paper, we propose a novel approach, the Relation and Semantic-sensitive Visual Grounding (ResVG) model, to address this issue. Firstly, we enhance the model's understanding of fine-grained semantics by injecting semantic prior information derived from text queries into the model. This is achieved by leveraging text-to-image generation models to produce images representing the semantic attributes of target objects described in queries. Secondly, we tackle the lack of training samples with multiple distractions by introducing a relation-sensitive data augmentation method. This method generates additional training data by synthesizing images containing multiple objects of the same category and pseudo queries based on their spatial relationships. The proposed ReSVG model significantly improves the model's ability to comprehend both object semantics and spatial relations, leading to enhanced performance in visual grounding tasks, particularly in scenarios with multiple-instance distractions. We conduct extensive experiments to validate the effectiveness of our methods on five datasets. Code is available at https://github.com/minghangz/ResVG.

Wenhao Li 0011, Qiangchang Wang, Peng Zhao 0016, Yilong Yin

Few-Shot Learning (FSL) aims to recognize new classes with limited labeled data. Recent studies have attempted to address the challenge of rare samples with textual prompts to modulate visual features. However, they usually struggle to capture complex semantic relationships between textual and visual features. Moreover, vanilla self-attention is heavily affected by useless information in images, severely constraining the potential of semantic priors in FSL due to the confusion of numerous irrelevant tokens during interaction. To address these aforementioned issues, a K-NN Transformer with Pyramid Prompts (KTPP) is proposed to select discriminative information with K-NN Context Attention (KCA) and adaptively modulate visual features with Pyramid Cross-modal Prompts (PCP). First, for each token, the KCA only selects the K most relevant tokens to compute the self-attention matrix and incorporates the mean of all tokens as the context prompt to provide the global context in three cascaded stages. As a result, irrelevant tokens can be progressively suppressed. Secondly, pyramid prompts are introduced in the PCP to emphasize visual features via interactions between text-based class-aware prompts and multi-scale visual features. This allows the ViT to dynamically adjust the importance weights of visual features based on rich semantic information at different scales, making models robust to spatial variations. Finally, augmented visual features and class-aware prompts are interacted via the KCA to extract class-specific features. Consequently, our model further enhances noise-free visual representations via deep cross-modal interactions, extracting generalized visual representation in scenarios with few labeled samples. Extensive experiments on four benchmark datasets demonstrate significant gains over the state-of-the-art methods, especially for the 1-shot task with 2.28% improvement on average due to semantically enhanced visual representations.

Ke Zhu, Liang Zhao, Zheng Ge, Xiangyu Zhang 0005

This paper makes the first attempt towards unsupervised preference alignment in Vision-Language Models (VLMs). We generate chosen and rejected responses with regard to the original and augmented image pairs, and conduct preference alignment with direct preference optimization. It is based on a core idea: properly designed augmentation to the image input will induce VLM to generate false but hard negative responses, which helps the model to learn from and produce more robust and powerful answers. The whole pipeline no longer hinges on supervision from GPT-4 or human involvement during alignment, and is highly efficient with few lines of code. With only 8k randomly sampled unsupervised data, it achieves 90% relative score to GPT-4 on complex reasoning in LLaVA-Bench, and improves LLaVA-7B/13B by 6.7%/5.6% score on complex multi-modal benchmark MM-Vet. Visualizations shows its improved ability to align with user-intentions. A series of ablations are firmly conducted to reveal the latent mechanism of the approach, which also indicates its potential towards further scaling.

Haonan Zheng 0001, Xinyang Deng, Wen Jiang 0002, Wenrui Li 0001

With Vision-Language Pre-training (VLP) models demonstrating powerful multimodal interaction capabilities, the application scenarios of neural networks are no longer confined to unimodal domains but have expanded to more complex multimodal V+L downstream tasks. The security vulnerabilities of unimodal models have been extensively examined, whereas those of VLP models remain challenging. We note that in CV models, the understanding of images comes from annotated information, while VLP models are designed to learn image representations directly from raw text. Motivated by this discrepancy, we developed the Feature Guidance Attack (FGA), a novel method that uses text representations to direct the perturbation of clean images, resulting in the generation of adversarial images. FGA is orthogonal to many advanced attack strategies in the unimodal domain, facilitating the direct application of rich research findings from the unimodal to the multimodal scenario. By appropriately introducing text attack into FGA, we construct Feature Guidance with Text Attack (FGA-T). Through the interaction of attacking two modalities, FGA-T achieves superior attack effects against VLP models. Moreover, incorporating data augmentation and momentum mechanisms significantly improves the black-box transferability of FGA-T. Our method demonstrates stable and effective attack capabilities across various datasets, downstream tasks, and both black-box and white-box settings, offering a unified baseline for exploring the robustness of VLP models.

Yonghao Liu 0001, Lan Huang 0002, Bowen Cao, Ximing Li 0002, Fausto Giunchiglia, Xiaoyue Feng, Renchu Guan

Graphs, as a fundamental data structure, have proven efficacy in modeling complex relationships between objects and are therefore found in wide web applications. Graph classification is an essential task in graph data analysis, which can effectively assist in extracting information and mining content from the web. Recently, few-shot graph classification, a more realistic and challenging task, has garnered great research interest. Existing few-shot graph classification models are all supervised, assuming abundant labeled data in base classes for meta-training. However, sufficient annotation is often challenging to obtain in practice due to high costs or demand for expertise. Moreover, they commonly adopt complicated meta-learning algorithms via episodic training to transfer prior knowledge from base classes. To break free from these constraints, in this paper, we propose a simple yet effective approach named SMART for unsupervised few-shot graph classification without using any labeled data. SMART employs transfer learning philosophy instead of the previously prevailing meta-learning paradigm, avoiding the need for sophisticated meta-learning algorithms. Additionally, we adopt a novel mixup strategy to augment the original graph data and leverage unsupervised pretraining on these data to obtain the expressive graph encoder. We also utilize the prompt tuning technique to alleviate the overfitting and low fine-tuning efficiency caused by the limited support samples of novel classes. Extensive experimental results demonstrate the superiority of our proposed approach, significantly surpassing even leading supervised few-shot graph classification models. Our code is available here.

Peilin Zhou, You-Liang Huang, Yueqi Xie, Jingqi Gao, Shoujin Wang, Jae Boum Kim, Sunghun Kim 0001

Sequential recommender systems (SRS) are designed to predict users' future behaviors based on their historical interaction data. Recent research has increasingly utilized contrastive learning (CL) to leverage unsupervised signals to alleviate the data sparsity issue in SRS. In general, CL-based SRS first augments the raw sequential interaction data by using data augmentation strategies and employs a contrastive training scheme to enforce the representations of those sequences from the same raw interaction data to be similar. Despite the growing popularity of CL, data augmentation, as a basic component of CL, has not received sufficient attention. This raises the question: Is it possible to achieve superior recommendation results solely through data augmentation? To answer this question, we benchmark eight widely used data augmentation strategies, as well as state-of-the-art CL-based SRS methods, on four real-world datasets under both warm- and cold-start settings. Intriguingly, the conclusion drawn from our study is that, certain data augmentation strategies can achieve similar or even superior performance compared with some CL-based methods, demonstrating the potential to significantly alleviate the data sparsity issue with fewer computational overhead. We hope that our study can further inspire more fundamental studies on the key functional components of complex CL techniques. Our processed datasets and codes are available at https://github.com/AIM-SE/DA4Rec.

Shaowei Wei, Zhengwei Wu, Xin Li 0090, Qintong Wu, Zhiqiang Zhang 0012, Jun Zhou 0011, Lihong Gu, Jinjie Gu

Sequential recommendation methods play a pivotal role in modern recommendation systems. A key challenge lies in accurately modeling user preferences in the face of data sparsity. To tackle this challenge, recent methods leverage contrastive learning (CL) to derive self-supervision signals by maximizing the mutual information of two augmented views of the original user behavior sequence. Despite their effectiveness, CL-based methods encounter a limitation in fully exploiting self-supervision signals for users with limited behavior data, as users with extensive behaviors naturally offer more information. To address this problem, we introduce a novel learning paradigm, named Online Self-Supervised Self-distillation for Sequential Recommendation (S4Rec), effectively bridging the gap between self-supervised learning and self-distillation methods. Specifically, we employ online clustering to proficiently group users by their distinct latent intents. Additionally, an adversarial learning strategy is utilized to ensure that the clustering procedure is not affected by the behavior length factor. Subsequently, we employ self-distillation to facilitate the transfer of knowledge from users with extensive behaviors (teachers) to users with limited behaviors (students). Experiments conducted on four real-world datasets validate the effectiveness of the proposed method1.

Dan Zhang, Yangliao Geng, Wenwen Gong, Zhongang Qi, Zhiyu Chen 0014, Xing Tang 0007, Ying Shan, Yuxiao Dong, Jie Tang 0001

Self-supervised learning (SSL) has recently achieved great success in mining the user-item interactions for collaborative filtering. As a major paradigm, contrastive learning (CL) based SSL helps address data sparsity in Web platforms by contrasting the embeddings between raw and augmented data. However, existing CL-based methods mostly focus on contrasting in a batch-wise way, failing to exploit potential regularity in the feature dimension. This leads to redundant solutions during the representation learning of users and items. In this work, we investigate how to employ both batch-wise CL (BCL) and feature-wise CL (FCL) for recommendation. We theoretically analyze the relation between BCL and FCL, and find that combining BCL and FCL helps eliminate redundant solutions but never misses an optimal solution. We propose a dual contrastive learning recommendation framework---RecDCL. In RecDCL, the FCL objective is designed to eliminate redundant solutions on user-item positive pairs and to optimize the uniform distributions within users and items using a polynomial kernel for driving the representations to be orthogonal; The BCL objective is utilized to generate contrastive embeddings on output vectors for enhancing the robustness of the representations. Extensive experiments on four widely-used benchmarks and one industry dataset demonstrate that RecDCL can consistently outperform the state-of-the-art GNNs-based and SSL-based models (with an improvement of up to 5.65% in terms of Recall@20). The source code is publicly available https://github.com/THUDM/RecDCL

Yu Wang 0176, Tongya Zheng, Yuxuan Liang 0002, Shunyu Liu 0001, Mingli Song

Human trajectory data produced by daily mobile devices has proven its usefulness in various substantial fields such as urban planning and epidemic prevention. In terms of the individual privacy concern, human trajectory simulation has attracted increasing attention from researchers, targeting at offering numerous realistic mobility data for downstream tasks. Nevertheless, the prevalent issue of data scarcity undoubtedly degrades the reliability of existing deep learning models. In this paper, we are motivated to explore the intriguing problem of mobility transfer across cities, grasping the universal patterns of human trajectories to augment the powerful Transformer with external mobility data. There are two crucial challenges arising in the knowledge transfer across cities: 1) how to transfer the Transformer to adapt for domain heterogeneity; 2) how to calibrate the Transformer to adapt for subtly different long-tail frequency distributions of locations. To address these challenges, we have tailored a Cross-city mObiLity trAnsformer (COLA) with a dedicated model-agnostic transfer framework by effectively transferring cross-city knowledge for human trajectory simulation. Firstly, COLA divides the Transformer into the private modules for city-specific characteristics and the shared modules for city-universal mobility patterns. Secondly, COLA leverages a lightweight yet effective post-hoc adjustment strategy for trajectory simulation, without disturbing the complex bi-level optimization of model-agnostic knowledge transfer. Extensive experiments of COLA compared to state-of-the-art single-city baselines and our implemented cross-city baselines have demonstrated its superiority and effectiveness. The code is available at https://github.com/Star607/Cross-city-Mobility-Transformer.

Jianghao Lin, Rong Shan, Chenxu Zhu, Kounianhua Du, Bo Chen 0023, Shigang Quan, Ruiming Tang, Yong Yu 0001, Weinan Zhang 0001

With large language models (LLMs) achieving remarkable breakthroughs in NLP domains, LLM-enhanced recommender systems have received much attention and have been actively explored currently. In this paper, we focus on adapting and empowering a pure large language model for zero-shot and few-shot recommendation tasks. First and foremost, we identify and formulate the lifelong sequential behavior incomprehension problem for LLMs in recommendation domains, i.e., LLMs fail to extract useful information from a textual context of long user behavior sequence, even if the length of context is far from reaching the context limitation of LLMs. To address such an issue and improve the recommendation performance of LLMs, we propose a novel framework, namely Retrieval enhanced Large Language models (ReLLa) for recommendation tasks in both zero-shot and few-shot settings. For zero-shot recommendation, we perform semantic user behavior retrieval (SUBR) to improve the data quality of testing samples, which greatly reduces the difficulty for LLMs to extract the essential knowledge from user behavior sequences. As for few-shot recommendation, we further design retrieval-enhanced instruction tuning (ReiT) by adopting SUBR as a data augmentation technique for training samples. Specifically, we develop a mixed training dataset consisting of both the original data samples and their retrieval-enhanced counterparts. We conduct extensive experiments on three real-world public datasets to demonstrate the superiority of ReLLa compared with existing baseline models, as well as its capability for lifelong sequential behavior comprehension. To be highlighted, with only less than 10% training samples, few-shot ReLLa can outperform traditional CTR models that are trained on the entire training set (e.g., DCNv2, DIN, SIM).

Jianghao Lin, Bo Chen 0023, Hangyu Wang, Yunjia Xi, Yanru Qu, Xinyi Dai, Kangning Zhang, Ruiming Tang, Yong Yu 0001, Weinan Zhang 0001

Click-through rate (CTR) prediction has become increasingly indispensable for various Internet applications. Traditional CTR models convert the multi-field categorical data into ID features via one-hot encoding, and extract the collaborative signals among features. Such a paradigm suffers from the problem of semantic information loss. Another line of research explores the potential of pretrained language models (PLMs) for CTR prediction by converting input data into textual sentences through hard prompt templates. Although semantic signals are preserved, they generally fail to capture the collaborative information (e.g., feature interactions, pure ID features), not to mention the unacceptable inference overhead brought by the huge model size. In this paper, we aim to model both the semantic knowledge and collaborative knowledge for accurate CTR estimation, and meanwhile address the inference inefficiency issue. To benefit from both worlds and close their gaps, we propose a novel model-agnostic framework (i.e., ClickPrompt), where we incorporate CTR models to generate interaction-aware soft prompts for PLMs. We design a prompt-augmented masked language modeling (PA-MLM) pretraining task, where PLM has to recover the masked tokens based on the language context, as well as the soft prompts generated by CTR model. The collaborative and semantic knowledge from ID and textual features would be explicitly aligned and interacted via the prompt interface. Then, we can either tune the CTR model with PLM for superior performance, or solely tune the CTR model without PLM for inference efficiency. Experiments on four real-world datasets validate the effectiveness of ClickPrompt compared with existing baselines.