Daily Papers

To extend the reinforcement learning post-training paradigm to omni-modal models for concurrently bolstering video-audio understanding and collaborative reasoning, we propose OmniJigsaw, a generic self-supervised framework built upon a temporal reordering proxy task. Centered on the chronological reconstruction of shuffled audio-visual clips, this paradigm strategically orchestrates visual and auditory signals to compel cross-modal integration through three distinct strategies: Joint Modality Integration, Sample-level Modality Selection, and Clip-level Modality Masking. Recognizing that the efficacy of such proxy tasks is fundamentally tied to puzzle quality, we design a two-stage coarse-to-fine data filtering pipeline, which facilitates the efficient adaptation of OmniJigsaw to massive unannotated omni-modal data. Our analysis reveals a ``bi-modal shortcut phenomenon'' in joint modality integration and demonstrates that fine-grained clip-level modality masking mitigates this issue while outperforming sample-level modality selection. Extensive evaluations on 15 benchmarks show substantial gains in video, audio, and collaborative reasoning, validating OmniJigsaw as a scalable paradigm for self-supervised omni-modal learning.

Masked modeling has emerged as a powerful self-supervised learning framework, but existing methods largely rely on random masking, disregarding the structural properties of different modalities. In this work, we introduce structured noise-based masking, a simple yet effective approach that naturally aligns with the spatial, temporal, and spectral characteristics of video and audio data. By filtering white noise into distinct color noise distributions, we generate structured masks that preserve modality-specific patterns without requiring handcrafted heuristics or access to the data. Our approach improves the performance of masked video and audio modeling frameworks without any computational overhead. Extensive experiments demonstrate that structured noise masking achieves consistent improvement over random masking for standard and advanced masked modeling methods, highlighting the importance of modality-aware masking strategies for representation learning.

Multimodal learning seeks to combine data from multiple input sources to enhance the performance of different downstream tasks. In real-world scenarios, performance can degrade substantially if some input modalities are missing. Existing methods that can handle missing modalities involve custom training or adaptation steps for each input modality combination. These approaches are either tied to specific modalities or become computationally expensive as the number of input modalities increases. In this paper, we propose Masked Modality Projection (MMP), a method designed to train a single model that is robust to any missing modality scenario. We achieve this by randomly masking a subset of modalities during training and learning to project available input modalities to estimate the tokens for the masked modalities. This approach enables the model to effectively learn to leverage the information from the available modalities to compensate for the missing ones, enhancing missing modality robustness. We conduct a series of experiments with various baseline models and datasets to assess the effectiveness of this strategy. Experiments demonstrate that our approach improves robustness to different missing modality scenarios, outperforming existing methods designed for missing modalities or specific modality combinations.

Large pre-trained vision-language models, such as CLIP, have demonstrated state-of-the-art performance across a wide range of image classification tasks, without requiring retraining. Few-shot CLIP is competitive with existing specialized architectures that were trained on the downstream tasks. Recent research demonstrates that the performance of CLIP can be further improved using lightweight adaptation approaches. However, previous methods adapt different modalities of the CLIP model individually, ignoring the interactions and relationships between visual and textual representations. In this work, we propose Multi-Modal Adapter, an approach for Multi-Modal adaptation of CLIP. Specifically, we add a trainable Multi-Head Attention layer that combines text and image features to produce an additive adaptation of both. Multi-Modal Adapter demonstrates improved generalizability, based on its performance on unseen classes compared to existing adaptation methods. We perform additional ablations and investigations to validate and interpret the proposed approach.

Recent approaches have shown that large-scale vision-language models such as CLIP can improve semantic segmentation performance. These methods typically aim for pixel-level vision-language alignment, but often rely on low resolution image features from CLIP, resulting in class ambiguities along boundaries. Moreover, the global scene representations in CLIP text embeddings do not directly correlate with the local and detailed pixel-level features, making meaningful alignment more difficult. To address these limitations, we introduce MTA-CLIP, a novel framework employing mask-level vision-language alignment. Specifically, we first propose Mask-Text Decoder that enhances the mask representations using rich textual data with the CLIP language model. Subsequently, it aligns mask representations with text embeddings using Mask-to-Text Contrastive Learning. Furthermore, we introduce MaskText Prompt Learning, utilizing multiple context-specific prompts for text embeddings to capture diverse class representations across masks. Overall, MTA-CLIP achieves state-of-the-art, surpassing prior works by an average of 2.8% and 1.3% on on standard benchmark datasets, ADE20k and Cityscapes, respectively.

Large-scale pre-trained image-text models demonstrate remarkable versatility across diverse tasks, benefiting from their robust representational capabilities and effective multimodal alignment. We extend the application of these models, specifically CLIP, to the domain of sound source localization. Unlike conventional approaches, we employ the pre-trained CLIP model without explicit text input, relying solely on the audio-visual correspondence. To this end, we introduce a framework that translates audio signals into tokens compatible with CLIP's text encoder, yielding audio-driven embeddings. By directly using these embeddings, our method generates audio-grounded masks for the provided audio, extracts audio-grounded image features from the highlighted regions, and aligns them with the audio-driven embeddings using the audio-visual correspondence objective. Our findings suggest that utilizing pre-trained image-text models enable our model to generate more complete and compact localization maps for the sounding objects. Extensive experiments show that our method outperforms state-of-the-art approaches by a significant margin.

We propose a pre-training strategy called Multi-modal Multi-task Masked Autoencoders (MultiMAE). It differs from standard Masked Autoencoding in two key aspects: I) it can optionally accept additional modalities of information in the input besides the RGB image (hence "multi-modal"), and II) its training objective accordingly includes predicting multiple outputs besides the RGB image (hence "multi-task"). We make use of masking (across image patches and input modalities) to make training MultiMAE tractable as well as to ensure cross-modality predictive coding is indeed learned by the network. We show this pre-training strategy leads to a flexible, simple, and efficient framework with improved transfer results to downstream tasks. In particular, the same exact pre-trained network can be flexibly used when additional information besides RGB images is available or when no information other than RGB is available - in all configurations yielding competitive to or significantly better results than the baselines. To avoid needing training datasets with multiple modalities and tasks, we train MultiMAE entirely using pseudo labeling, which makes the framework widely applicable to any RGB dataset. The experiments are performed on multiple transfer tasks (image classification, semantic segmentation, depth estimation) and datasets (ImageNet, ADE20K, Taskonomy, Hypersim, NYUv2). The results show an intriguingly impressive capability by the model in cross-modal/task predictive coding and transfer.

The popular CLIP model displays impressive zero-shot capabilities thanks to its seamless interaction with arbitrary text prompts. However, its lack of spatial awareness makes it unsuitable for dense computer vision tasks, e.g., semantic segmentation, without an additional fine-tuning step that often uses annotations and can potentially suppress its original open-vocabulary properties. Meanwhile, self-supervised representation methods have demonstrated good localization properties without human-made annotations nor explicit supervision. In this work, we take the best of both worlds and propose a zero-shot open-vocabulary semantic segmentation method, which does not require any annotations. We propose to locally improve dense MaskCLIP features, computed with a simple modification of CLIP's last pooling layer, by integrating localization priors extracted from self-supervised features. By doing so, we greatly improve the performance of MaskCLIP and produce smooth outputs. Moreover, we show that the used self-supervised feature properties can directly be learnt from CLIP features therefore allowing us to obtain the best results with a single pass through CLIP model. Our method CLIP-DINOiser needs only a single forward pass of CLIP and two light convolutional layers at inference, no extra supervision nor extra memory and reaches state-of-the-art results on challenging and fine-grained benchmarks such as COCO, Pascal Context, Cityscapes and ADE20k. The code to reproduce our results is available at https://github.com/wysoczanska/clip_dinoiser.

Pre-trained multi-modal Vision-Language Models like CLIP are widely used off-the-shelf for a variety of applications. In this paper, we show that the common practice of individually exploiting the text or image encoders of these powerful multi-modal models is highly suboptimal for intra-modal tasks like image-to-image retrieval. We argue that this is inherently due to the CLIP-style inter-modal contrastive loss that does not enforce any intra-modal constraints, leading to what we call intra-modal misalignment. To demonstrate this, we leverage two optimization-based modality inversion techniques that map representations from their input modality to the complementary one without any need for auxiliary data or additional trained adapters. We empirically show that, in the intra-modal tasks of image-to-image and text-to-text retrieval, approaching these tasks inter-modally significantly improves performance with respect to intra-modal baselines on more than fifteen datasets. Additionally, we demonstrate that approaching a native inter-modal task (e.g. zero-shot image classification) intra-modally decreases performance, further validating our findings. Finally, we show that incorporating an intra-modal term in the pre-training objective or narrowing the modality gap between the text and image feature embedding spaces helps reduce the intra-modal misalignment. The code is publicly available at: https://github.com/miccunifi/Cross-the-Gap.

Contrastive language image pre-training (CLIP) is an essential component of building modern vision-language foundation models. While CLIP demonstrates remarkable zero-shot performance on downstream tasks, the multi-modal feature spaces still suffer from a modality gap, which is a gap between image and text feature clusters and limits downstream task performance. Although existing works attempt to address the modality gap by modifying pre-training or fine-tuning, they struggle with heavy training costs with large datasets or degradations of zero-shot performance. This paper presents CLIP-Refine, a post-pre-training method for CLIP models at a phase between pre-training and fine-tuning. CLIP-Refine aims to align the feature space with 1 epoch training on small image-text datasets without zero-shot performance degradations. To this end, we introduce two techniques: random feature alignment (RaFA) and hybrid contrastive-distillation (HyCD). RaFA aligns the image and text features to follow a shared prior distribution by minimizing the distance to random reference vectors sampled from the prior. HyCD updates the model with hybrid soft labels generated by combining ground-truth image-text pair labels and outputs from the pre-trained CLIP model. This contributes to achieving both maintaining the past knowledge and learning new knowledge to align features. Our extensive experiments with multiple classification and retrieval tasks show that CLIP-Refine succeeds in mitigating the modality gap and improving the zero-shot performance.

Contrastive Language--Image Pre-training (CLIP) has manifested remarkable improvements in zero-shot classification and cross-modal vision-language tasks. Yet, from a geometrical point of view, the CLIP embedding space has been found to have a pronounced modality gap. This gap renders the embedding space overly sparse and disconnected, with different modalities being densely distributed in distinct subregions of the hypersphere. In this work, we aim at answering two main questions: 1. Does sharing the parameter space between the multi-modal encoders reduce the modality gap? 2. Can the gap be mitigated by pushing apart the uni-modal embeddings via intra-modality separation? We design AlignCLIP, in order to answer these questions and show that answers to both questions are positive. Through extensive experiments, we show that AlignCLIP achieves noticeable enhancements in the cross-modal alignment of the embeddings, and thereby, reduces the modality gap, while maintaining the performance across several downstream evaluations, such as zero-shot image classification, zero-shot multi-modal retrieval and zero-shot semantic text similarity.

Recent advancements in generative models have revolutionized image generation and editing, making these tasks accessible to non-experts. This paper focuses on local image editing, particularly the task of adding new content to a loosely specified area. Existing methods often require a precise mask or a detailed description of the location, which can be cumbersome and prone to errors. We propose Click2Mask, a novel approach that simplifies the local editing process by requiring only a single point of reference (in addition to the content description). A mask is dynamically grown around this point during a Blended Latent Diffusion (BLD) process, guided by a masked CLIP-based semantic loss. Click2Mask surpasses the limitations of segmentation-based and fine-tuning dependent methods, offering a more user-friendly and contextually accurate solution. Our experiments demonstrate that Click2Mask not only minimizes user effort but also delivers competitive or superior local image manipulation results compared to SoTA methods, according to both human judgement and automatic metrics. Key contributions include the simplification of user input, the ability to freely add objects unconstrained by existing segments, and the integration potential of our dynamic mask approach within other editing methods.

Contrastive pre-training on image-text pairs, exemplified by CLIP, becomes a standard technique for learning multi-modal visual-language representations. Although CLIP has demonstrated remarkable performance, training it from scratch on noisy web-scale datasets is computationally demanding. On the other hand, mask-then-predict pre-training approaches, like Masked Image Modeling (MIM), offer efficient self-supervised learning for single-modal representations. This paper introduces Unmasked Token Alignment (UTA), a method that leverages existing CLIP models to further enhance its vision-language representations. UTA trains a Vision Transformer (ViT) by aligning unmasked visual tokens to the corresponding image tokens from a frozen CLIP vision encoder, which automatically aligns the ViT model with the CLIP text encoder. The pre-trained ViT can be directly applied for zero-shot evaluation even without training on image-text pairs. Compared to MIM approaches, UTA does not suffer from training-finetuning inconsistency and is much more training-efficient by avoiding using the extra [MASK] tokens. Extensive experimental results demonstrate that UTA can enhance CLIP models and outperform existing MIM methods on various uni- and multi-modal benchmarks. Code and models are available at https://github.com/jihaonew/UTA.

Vision-language pre-trained models have achieved impressive performance on various downstream tasks. However, their large model sizes hinder their utilization on platforms with limited computational resources. We find that directly using smaller pre-trained models and applying magnitude-based pruning on CLIP models leads to inflexibility and inferior performance. Recent efforts for VLP compression either adopt uni-modal compression metrics resulting in limited performance or involve costly mask-search processes with learnable masks. In this paper, we first propose the Module-wise Pruning Error (MoPE) metric, accurately assessing CLIP module importance by performance decline on cross-modal tasks. Using the MoPE metric, we introduce a unified pruning framework applicable to both pre-training and task-specific fine-tuning compression stages. For pre-training, MoPE-CLIP effectively leverages knowledge from the teacher model, significantly reducing pre-training costs while maintaining strong zero-shot capabilities. For fine-tuning, consecutive pruning from width to depth yields highly competitive task-specific models. Extensive experiments in two stages demonstrate the effectiveness of the MoPE metric, and MoPE-CLIP outperforms previous state-of-the-art VLP compression methods.

Pre-trained vision-language models~(VLMs) are the de-facto foundation models for various downstream tasks. However, scene text recognition methods still prefer backbones pre-trained on a single modality, namely, the visual modality, despite the potential of VLMs to serve as powerful scene text readers. For example, CLIP can robustly identify regular (horizontal) and irregular (rotated, curved, blurred, or occluded) text in images. With such merits, we transform CLIP into a scene text reader and introduce CLIP4STR, a simple yet effective STR method built upon image and text encoders of CLIP. It has two encoder-decoder branches: a visual branch and a cross-modal branch. The visual branch provides an initial prediction based on the visual feature, and the cross-modal branch refines this prediction by addressing the discrepancy between the visual feature and text semantics. To fully leverage the capabilities of both branches, we design a dual predict-and-refine decoding scheme for inference. We scale CLIP4STR in terms of the model size, pre-training data, and training data, achieving state-of-the-art performance on 11 STR benchmarks. Additionally, a comprehensive empirical study is provided to enhance the understanding of the adaptation of CLIP to STR. We believe our method establishes a simple yet strong baseline for future STR research with VLMs.

We present a simplified, task-agnostic multi-modal pre-training approach that can accept either video or text input, or both for a variety of end tasks. Existing pre-training are task-specific by adopting either a single cross-modal encoder that requires both modalities, limiting their use for retrieval-style end tasks or more complex multitask learning with two unimodal encoders, limiting early cross-modal fusion. We instead introduce new pretraining masking schemes that better mix across modalities (e.g. by forcing masks for text to predict the closest video embeddings) while also maintaining separability (e.g. unimodal predictions are sometimes required, without using all the input). Experimental results show strong performance across a wider range of tasks than any previous methods, often outperforming task-specific pre-training. Code is made available at https://github.com/pytorch/fairseq/tree/main/examples/MMPT.

Open-vocabulary semantic segmentation aims to segment an image into semantic regions according to text descriptions, which may not have been seen during training. Recent two-stage methods first generate class-agnostic mask proposals and then leverage pre-trained vision-language models, e.g., CLIP, to classify masked regions. We identify the performance bottleneck of this paradigm to be the pre-trained CLIP model, since it does not perform well on masked images. To address this, we propose to finetune CLIP on a collection of masked image regions and their corresponding text descriptions. We collect training data by mining an existing image-caption dataset (e.g., COCO Captions), using CLIP to match masked image regions to nouns in the image captions. Compared with the more precise and manually annotated segmentation labels with fixed classes (e.g., COCO-Stuff), we find our noisy but diverse dataset can better retain CLIP's generalization ability. Along with finetuning the entire model, we utilize the "blank" areas in masked images using a method we dub mask prompt tuning. Experiments demonstrate mask prompt tuning brings significant improvement without modifying any weights of CLIP, and it can further improve a fully finetuned model. In particular, when trained on COCO and evaluated on ADE20K-150, our best model achieves 29.6% mIoU, which is +8.5% higher than the previous state-of-the-art. For the first time, open-vocabulary generalist models match the performance of supervised specialist models in 2017 without dataset-specific adaptations.

Vision-Language Models like CLIP are extensively used for inter-modal tasks which involve both visual and text modalities. However, when the individual modality encoders are applied to inherently intra-modal tasks like image-to-image retrieval, their performance suffers from the intra-modal misalignment. In this paper we study intra-modal misalignment in CLIP with a focus on the role of the projectors that map pre-projection image and text embeddings into the shared embedding space. By analyzing the form of the cosine similarity applied to projected features, and its interaction with the contrastive CLIP loss, we show that there is an inter-modal operator responsible for aligning the two modalities during training, and a second, intra-modal operator that only enforces intra-modal normalization but does nothing to promote intra-modal alignment. Via spectral analysis of the inter-modal operator, we identify an approximately isotropic subspace in which the two modalities are well-aligned, as well as anisotropic directions specific to each modality. We demonstrate that this aligned subspace can be directly obtained from the projector weights and that removing the anisotropic directions improves intra-modal alignment. Our experiments on intra-modal retrieval and classification benchmarks show that our training-free method reduces intra-modal misalignment, greatly lowers latency, and outperforms existing approaches across multiple pre-trained CLIP-like models. The code is publicly available at: https://github.com/simomagi/IsoCLIP.

The synergy between generative and discriminative models receives growing attention. While discriminative Contrastive Language-Image Pre-Training (CLIP) excels in high-level semantics, it struggles with perceiving fine-grained visual details. Generally, to enhance representations, generative models take CLIP's visual features as conditions for reconstruction. However, the underlying principle remains underexplored. In this work, we empirically found that visually perfect generations are not always optimal for representation enhancement. The essence lies in effectively extracting fine-grained knowledge from generative models while mitigating irrelevant information. To explore critical factors, we delve into three aspects: (1) Conditioning mechanisms: We found that even a small number of local tokens can drastically reduce the difficulty of reconstruction, leading to collapsed training. We thus conclude that utilizing only global visual tokens as conditions is the most effective strategy. (2) Denoising configurations: We observed that end-to-end training introduces extraneous information. To address this, we propose a two-stage training strategy to prioritize learning useful visual knowledge. Additionally, we demonstrate that lightweight denoisers can yield remarkable improvements. (3) Generation paradigms: We explore both continuous and discrete denoisers with desirable outcomes, validating the versatility of our method. Through our in-depth explorations, we have finally arrived at an effective method, namely GenHancer, which consistently outperforms prior arts on the MMVP-VLM benchmark, e.g., 6.0% on OpenAICLIP. The enhanced CLIP can be further plugged into multimodal large language models for better vision-centric performance. All the models and codes are made publicly available.

CLIP is a discriminative model trained to align images and text in a shared embedding space. Due to its multimodal structure, it serves as the backbone of many generative pipelines, where a decoder is trained to map from the shared space back to images. In this work, we show that image synthesis is nevertheless possible using CLIP alone -- without any decoder, training, or fine-tuning. Our approach optimizes a frequency-aware implicit neural representation that encourages coarse-to-fine generation by stratifying frequencies across network layers. To stabilize this inverse mapping, we introduce adversarially robust initialization, a lightweight Orthogonal Procrustes projection to align local text and image embeddings, and a blending loss that anchors outputs to natural image statistics. Without altering CLIP's weights, this framework unlocks capabilities such as text-to-image generation, style transfer, and image reconstruction. These findings suggest that discriminative models may hold untapped generative potential, hidden in plain sight.

Recently, masked video modeling has been widely explored and significantly improved the model's understanding ability of visual regions at a local level. However, existing methods usually adopt random masking and follow the same reconstruction paradigm to complete the masked regions, which do not leverage the correlations between cross-modal content. In this paper, we present Mask for Semantics Completion (MASCOT) based on semantic-based masked modeling. Specifically, after applying attention-based video masking to generate high-informed and low-informed masks, we propose Informed Semantics Completion to recover masked semantics information. The recovery mechanism is achieved by aligning the masked content with the unmasked visual regions and corresponding textual context, which makes the model capture more text-related details at a patch level. Additionally, we shift the emphasis of reconstruction from irrelevant backgrounds to discriminative parts to ignore regions with low-informed masks. Furthermore, we design dual-mask co-learning to incorporate video cues under different masks and learn more aligned video representation. Our MASCOT performs state-of-the-art performance on four major text-video retrieval benchmarks, including MSR-VTT, LSMDC, ActivityNet, and DiDeMo. Extensive ablation studies demonstrate the effectiveness of the proposed schemes.

Recent work has studied text-to-audio synthesis using large amounts of paired text-audio data. However, audio recordings with high-quality text annotations can be difficult to acquire. In this work, we approach text-to-audio synthesis using unlabeled videos and pretrained language-vision models. We propose to learn the desired text-audio correspondence by leveraging the visual modality as a bridge. We train a conditional diffusion model to generate the audio track of a video, given a video frame encoded by a pretrained contrastive language-image pretraining (CLIP) model. At test time, we first explore performing a zero-shot modality transfer and condition the diffusion model with a CLIP-encoded text query. However, we observe a noticeable performance drop with respect to image queries. To close this gap, we further adopt a pretrained diffusion prior model to generate a CLIP image embedding given a CLIP text embedding. Our results show the effectiveness of the proposed method, and that the pretrained diffusion prior can reduce the modality transfer gap. While we focus on text-to-audio synthesis, the proposed model can also generate audio from image queries, and it shows competitive performance against a state-of-the-art image-to-audio synthesis model in a subjective listening test. This study offers a new direction of approaching text-to-audio synthesis that leverages the naturally-occurring audio-visual correspondence in videos and the power of pretrained language-vision models.

Mixed modality search -- retrieving information across a heterogeneous corpus composed of images, texts, and multimodal documents -- is an important yet underexplored real-world application. In this work, we investigate how contrastive vision-language models, such as CLIP, perform on the mixed modality search task. Our analysis reveals a critical limitation: these models exhibit a pronounced modality gap in the embedding space, where image and text embeddings form distinct clusters, leading to intra-modal ranking bias and inter-modal fusion failure. To address this issue, we propose GR-CLIP, a lightweight post-hoc calibration method that removes the modality gap in CLIP's embedding space. Evaluated on MixBench -- the first benchmark specifically designed for mixed modality search -- GR-CLIP improves NDCG@10 by up to 26 percentage points over CLIP, surpasses recent vision-language generative embedding models by 4 percentage points, while using 75x less compute.

Pre-trained multi-modal models, such as CLIP, provide transferable embeddings and show promising results in diverse applications. However, the analysis of learned multi-modal embeddings is relatively unexplored, and the embedding transferability can be improved. In this work, we observe that CLIP holds separated embedding subspaces for two different modalities, and then we investigate it through the lens of uniformity-alignment to measure the quality of learned representation. Both theoretically and empirically, we show that CLIP retains poor uniformity and alignment even after fine-tuning. Such a lack of alignment and uniformity might restrict the transferability and robustness of embeddings. To this end, we devise a new fine-tuning method for robust representation equipping better alignment and uniformity. First, we propose a Geodesic Multi-Modal Mixup that mixes the embeddings of image and text to generate hard negative samples on the hypersphere. Then, we fine-tune the model on hard negatives as well as original negatives and positives with contrastive loss. Based on the theoretical analysis about hardness guarantee and limiting behavior, we justify the use of our method. Extensive experiments on retrieval, calibration, few- or zero-shot classification (under distribution shift), embedding arithmetic, and image captioning further show that our method provides transferable representations, enabling robust model adaptation on diverse tasks. Code: https://github.com/changdaeoh/multimodal-mixup

Multimodal learning has recently gained significant popularity, demonstrating impressive performance across various zero-shot classification tasks and a range of perceptive and generative applications. Models such as Contrastive Language-Image Pretraining (CLIP) are designed to bridge different modalities, such as images and text, by learning a shared representation space through contrastive learning. Despite their success, the working mechanisms underlying multimodal learning are not yet well understood. Notably, these models often exhibit a modality gap, where different modalities occupy distinct regions within the shared representation space. In this work, we conduct an in-depth analysis of the emergence of modality gap by characterizing the gradient flow learning dynamics. Specifically, we identify the critical roles of mismatched data pairs and a learnable temperature parameter in causing and perpetuating the modality gap during training. Furthermore, our theoretical insights are validated through experiments on practical CLIP models. These findings provide principled guidance for mitigating the modality gap, including strategies such as appropriate temperature scheduling and modality swapping. Additionally, we demonstrate that closing the modality gap leads to improved performance on tasks such as image-text retrieval.

Segment Anything Model (SAM) has recently shown its powerful effectiveness in visual segmentation tasks. However, there is less exploration concerning how SAM works on audio-visual tasks, such as visual sound localization and segmentation. In this work, we propose a simple yet effective audio-visual localization and segmentation framework based on the Segment Anything Model, namely AV-SAM, that can generate sounding object masks corresponding to the audio. Specifically, our AV-SAM simply leverages pixel-wise audio-visual fusion across audio features and visual features from the pre-trained image encoder in SAM to aggregate cross-modal representations. Then, the aggregated cross-modal features are fed into the prompt encoder and mask decoder to generate the final audio-visual segmentation masks. We conduct extensive experiments on Flickr-SoundNet and AVSBench datasets. The results demonstrate that the proposed AV-SAM can achieve competitive performance on sounding object localization and segmentation.

In this paper, we tackle an emerging computer vision task, open-vocabulary universal image segmentation, that aims to perform semantic/instance/panoptic segmentation (background semantic labeling + foreground instance segmentation) for arbitrary categories of text-based descriptions in inference time. We first build a baseline method by directly adopting pre-trained CLIP models without finetuning or distillation. We then develop MaskCLIP, a Transformer-based approach with a MaskCLIP Visual Encoder, which is an encoder-only module that seamlessly integrates mask tokens with a pre-trained ViT CLIP model for semantic/instance segmentation and class prediction. MaskCLIP learns to efficiently and effectively utilize pre-trained partial/dense CLIP features within the MaskCLIP Visual Encoder that avoids the time-consuming student-teacher training process. MaskCLIP outperforms previous methods for semantic/instance/panoptic segmentation on ADE20K and PASCAL datasets. We show qualitative illustrations for MaskCLIP with online custom categories. Project website: https://maskclip.github.io.

Image captioning aims at generating descriptive and meaningful textual descriptions of images, enabling a broad range of vision-language applications. Prior works have demonstrated that harnessing the power of Contrastive Image Language Pre-training (CLIP) offers a promising approach to achieving zero-shot captioning, eliminating the need for expensive caption annotations. However, the widely observed modality gap in the latent space of CLIP harms the performance of zero-shot captioning by breaking the alignment between paired image-text features. To address this issue, we conduct an analysis on the CLIP latent space which leads to two findings. Firstly, we observe that the CLIP's visual feature of image subregions can achieve closer proximity to the paired caption due to the inherent information loss in text descriptions. In addition, we show that the modality gap between a paired image-text can be empirically modeled as a zero-mean Gaussian distribution. Motivated by the findings, we propose a novel zero-shot image captioning framework with text-only training to reduce the modality gap. In particular, we introduce a subregion feature aggregation to leverage local region information, which produces a compact visual representation for matching text representation. Moreover, we incorporate a noise injection and CLIP reranking strategy to boost captioning performance. We also extend our framework to build a zero-shot VQA pipeline, demonstrating its generality. Through extensive experiments on common captioning and VQA datasets such as MSCOCO, Flickr30k and VQAV2, we show that our method achieves remarkable performance improvements. Code is available at https://github.com/Artanic30/MacCap.

Multi-modal contrastive models such as CLIP achieve state-of-the-art performance in zero-shot classification by embedding input images and texts on a joint representational space. Recently, a modality gap has been reported in two-encoder contrastive models like CLIP, meaning that the image and text embeddings reside in disjoint areas of the latent space. Previous studies suggest that this gap exists due to 1) the cone effect, 2) mismatched pairs in the dataset, and 3) insufficient training. We show that, even when accounting for all these factors, and even when using the same modality, the contrastive loss actually creates a gap during training. As a result, We propose that the modality gap is inherent to the two-encoder contrastive loss and rename it the contrastive gap. We present evidence that attributes this contrastive gap to low uniformity in CLIP space, resulting in embeddings that occupy only a small portion of the latent space. To close the gap, we adapt the uniformity and alignment properties of unimodal contrastive loss to the multi-modal setting and show that simply adding these terms to the CLIP loss distributes the embeddings more uniformly in the representational space, closing the gap. In our experiments, we show that the modified representational space achieves better performance than default CLIP loss in downstream tasks such as zero-shot image classification and multi-modal arithmetic.

Contrastive Language-Image Pre-training (CLIP) models have shown significant potential, particularly in zero-shot classification across diverse distribution shifts. Building on existing evaluations of overall classification robustness, this work aims to provide a more comprehensive assessment of CLIP by introducing several new perspectives. First, we investigate their robustness to variations in specific visual factors. Second, we assess two critical safety objectives--confidence uncertainty and out-of-distribution detection--beyond mere classification accuracy. Third, we evaluate the finesse with which CLIP models bridge the image and text modalities. Fourth, we extend our examination to 3D awareness in CLIP models, moving beyond traditional 2D image understanding. Finally, we explore the interaction between vision and language encoders within modern large multimodal models (LMMs) that utilize CLIP as the visual backbone, focusing on how this interaction impacts classification robustness. In each aspect, we consider the impact of six factors on CLIP models: model architecture, training distribution, training set size, fine-tuning, contrastive loss, and test-time prompts. Our study uncovers several previously unknown insights into CLIP. For instance, the architecture of the visual encoder in CLIP plays a significant role in their robustness against 3D corruption. CLIP models tend to exhibit a bias towards shape when making predictions. Moreover, this bias tends to diminish after fine-tuning on ImageNet. Vision-language models like LLaVA, leveraging the CLIP vision encoder, could exhibit benefits in classification performance for challenging categories over CLIP alone. Our findings are poised to offer valuable guidance for enhancing the robustness and reliability of CLIP models.

In order to appropriately filter multi-modality data sets on a web-scale, it becomes crucial to employ suitable filtering methods to boost performance and reduce training costs. For instance, LAION papers employs the CLIP score filter to select data with CLIP scores surpassing a certain threshold. On the other hand, T-MARS achieves high-quality data filtering by detecting and masking text within images and then filtering by CLIP score. Through analyzing the dataset, we observe a significant proportion of redundant information, such as numbers, present in the textual content. Our experiments on a subset of the data unveil the profound impact of these redundant elements on the CLIP scores. A logical approach would involve reevaluating the CLIP scores after eliminating these influences. Experimentally, our text-based CLIP filter outperforms the top-ranked method on the ``small scale" of DataComp (a data filtering benchmark) on ImageNet distribution shifts, achieving a 3.6% performance improvement. The results also demonstrate that our proposed text-masked filter outperforms the original CLIP score filter when selecting the top 40% of the data. The impact of numbers on CLIP and their handling provide valuable insights for improving the effectiveness of CLIP training, including language rewrite techniques.

Identifying highlight moments of raw video materials is crucial for improving the efficiency of editing videos that are pervasive on internet platforms. However, the extensive work of manually labeling footage has created obstacles to applying supervised methods to videos of unseen categories. The absence of an audio modality that contains valuable cues for highlight detection in many videos also makes it difficult to use multimodal strategies. In this paper, we propose a novel model with cross-modal perception for unsupervised highlight detection. The proposed model learns representations with visual-audio level semantics from image-audio pair data via a self-reconstruction task. To achieve unsupervised highlight detection, we investigate the latent representations of the network and propose the representation activation sequence learning (RASL) module with k-point contrastive learning to learn significant representation activations. To connect the visual modality with the audio modality, we use the symmetric contrastive learning (SCL) module to learn the paired visual and audio representations. Furthermore, an auxiliary task of masked feature vector sequence (FVS) reconstruction is simultaneously conducted during pretraining for representation enhancement. During inference, the cross-modal pretrained model can generate representations with paired visual-audio semantics given only the visual modality. The RASL module is used to output the highlight scores. The experimental results show that the proposed framework achieves superior performance compared to other state-of-the-art approaches.

Contrastive Language-Image Pre-training (CLIP) has made a remarkable breakthrough in open-vocabulary zero-shot image recognition. Many recent studies leverage the pre-trained CLIP models for image-level classification and manipulation. In this paper, we wish examine the intrinsic potential of CLIP for pixel-level dense prediction, specifically in semantic segmentation. To this end, with minimal modification, we show that MaskCLIP yields compelling segmentation results on open concepts across various datasets in the absence of annotations and fine-tuning. By adding pseudo labeling and self-training, MaskCLIP+ surpasses SOTA transductive zero-shot semantic segmentation methods by large margins, e.g., mIoUs of unseen classes on PASCAL VOC/PASCAL Context/COCO Stuff are improved from 35.6/20.7/30.3 to 86.1/66.7/54.7. We also test the robustness of MaskCLIP under input corruption and evaluate its capability in discriminating fine-grained objects and novel concepts. Our finding suggests that MaskCLIP can serve as a new reliable source of supervision for dense prediction tasks to achieve annotation-free segmentation. Source code is available at https://github.com/chongzhou96/MaskCLIP.

This study aims to construct an audio-video generative model with minimal computational cost by leveraging pre-trained single-modal generative models for audio and video. To achieve this, we propose a novel method that guides single-modal models to cooperatively generate well-aligned samples across modalities. Specifically, given two pre-trained base diffusion models, we train a lightweight joint guidance module to adjust scores separately estimated by the base models to match the score of joint distribution over audio and video. We show that this guidance can be computed using the gradient of the optimal discriminator, which distinguishes real audio-video pairs from fake ones independently generated by the base models. Based on this analysis, we construct a joint guidance module by training this discriminator. Additionally, we adopt a loss function to stabilize the discriminator's gradient and make it work as a noise estimator, as in standard diffusion models. Empirical evaluations on several benchmark datasets demonstrate that our method improves both single-modal fidelity and multimodal alignment with relatively few parameters. The code is available at: https://github.com/SonyResearch/MMDisCo.

Multimodal Large Language Models have demonstrated impressive capabilities across tasks, yet they often exhibit difficulty in distinguishing task-relevant from irrelevant signals -- particularly in tasks like Visual Question Answering -- which can lead to susceptibility to misleading or spurious inputs. We refer to this broader limitation as the Cross-Modality Competency Problem -- the model's inability to fairly evaluate all modalities. This vulnerability becomes more evident in modality-specific tasks -- such as image classification or pure text question answering -- where models are expected to rely solely on one modality. In such tasks, spurious information from irrelevant modalities often leads to significant performance degradation. We refer to this failure as Modality Interference, which serves as a concrete and measurable instance of the cross-modality competency problem, and we further design a perturbation-based causal diagnostic experiment to verify and quantify this problem. To mitigate modality interference, we propose a novel framework to finetune MLLMs, including perturbation-based data augmentations with both heuristic perturbations and adversarial perturbations, and a consistency regularization strategy applying on model outputs with original and perturbed inputs. Experiments on multiple benchmark datasets (image-heavy, text-heavy and multimodal tasks) and multiple model families with different scales demonstrate significant improvements in robustness and cross-modality competency, indicating our method's effectiveness in boosting unimodal reasoning ability while enhancing performance on multimodal tasks.

Multi-modal learning has become increasingly popular due to its ability to leverage information from different data sources (e.g., text and images) to improve the model performance. Recently, CLIP has emerged as an effective approach that employs vision-language contrastive pretraining to learn joint image and text representations and exhibits remarkable performance in zero-shot learning and text-guided natural image generation. Despite the huge practical success of CLIP, its theoretical understanding remains elusive. In this paper, we formally study transferrable representation learning underlying CLIP and demonstrate how features from different modalities get aligned. We also analyze its zero-shot transfer performance on the downstream tasks. Inspired by our analysis, we propose a new CLIP-type approach, which achieves better performance than CLIP and other state-of-the-art methods on benchmark datasets.

Recently, the rise of large-scale vision-language pretrained models like CLIP, coupled with the technology of Parameter-Efficient FineTuning (PEFT), has captured substantial attraction in video action recognition. Nevertheless, prevailing approaches tend to prioritize strong supervised performance at the expense of compromising the models' generalization capabilities during transfer. In this paper, we introduce a novel Multimodal, Multi-task CLIP adapting framework named \name to address these challenges, preserving both high supervised performance and robust transferability. Firstly, to enhance the individual modality architectures, we introduce multimodal adapters to both the visual and text branches. Specifically, we design a novel visual TED-Adapter, that performs global Temporal Enhancement and local temporal Difference modeling to improve the temporal representation capabilities of the visual encoder. Moreover, we adopt text encoder adapters to strengthen the learning of semantic label information. Secondly, we design a multi-task decoder with a rich set of supervisory signals to adeptly satisfy the need for strong supervised performance and generalization within a multimodal framework. Experimental results validate the efficacy of our approach, demonstrating exceptional performance in supervised learning while maintaining strong generalization in zero-shot scenarios.

The computer vision domain has greatly benefited from an abundance of data across many modalities to improve on various visual tasks. Recently, there has been a lot of focus on self-supervised pre-training methods through Masked Autoencoders (MAE) he2022masked,bachmann2022multimae, usually used as a first step before optimizing for a downstream task, such as classification or regression. This is very useful as it doesn't require any manually labeled data. In this work, we introduce Probabilistic Hyper-Graphs using Masked Autoencoders (PHG-MAE): a novel model that unifies the classical work on neural graphs leordeanu2021semi with the modern approach of masked autoencoders under a common theoretical framework. Through random masking of entire modalities, not just patches, the model samples from the distribution of hyper-edges on each forward pass. Additionally, the model adapts the standard MAE algorithm by combining pre-training and fine-tuning into a single training loop. Moreover, our approach enables the creation of inference-time ensembles which, through aggregation, boost the final prediction performance and consistency. Lastly, we show that we can apply knowledge distillation on top of the ensembles with little loss in performance, even with models that have fewer than 1M parameters. While our work mostly focuses on outdoor UAV scenes that contain multiple world interpretations and modalities, the same steps can be followed in other similar domains, such as autonomous driving or indoor robotics. In order to streamline the process of integrating external pre-trained experts for computer vision multi-modal multi-task learning (MTL) scenarios, we developed a data-pipeline software. Using this tool, we have created and released a fully-automated extension of the Dronescapes dataset. All the technical details, code and reproduction steps are publicly released.

With the rapid growth in deepfake video content, we require improved and generalizable methods to detect them. Most existing detection methods either use uni-modal cues or rely on supervised training to capture the dissonance between the audio and visual modalities. While the former disregards the audio-visual correspondences entirely, the latter predominantly focuses on discerning audio-visual cues within the training corpus, thereby potentially overlooking correspondences that can help detect unseen deepfakes. We present Audio-Visual Feature Fusion (AVFF), a two-stage cross-modal learning method that explicitly captures the correspondence between the audio and visual modalities for improved deepfake detection. The first stage pursues representation learning via self-supervision on real videos to capture the intrinsic audio-visual correspondences. To extract rich cross-modal representations, we use contrastive learning and autoencoding objectives, and introduce a novel audio-visual complementary masking and feature fusion strategy. The learned representations are tuned in the second stage, where deepfake classification is pursued via supervised learning on both real and fake videos. Extensive experiments and analysis suggest that our novel representation learning paradigm is highly discriminative in nature. We report 98.6% accuracy and 99.1% AUC on the FakeAVCeleb dataset, outperforming the current audio-visual state-of-the-art by 14.9% and 9.9%, respectively.

This paper presents that the masked-modeling principle driving the success of large foundational vision models can be effectively applied to audio by making predictions in a latent space. We introduce Audio-based Joint-Embedding Predictive Architecture (A-JEPA), a simple extension method for self-supervised learning from the audio spectrum. Following the design of I-JEPA, our A-JEPA encodes visible audio spectrogram patches with a curriculum masking strategy via context encoder, and predicts the representations of regions sampled at well-designed locations. The target representations of those regions are extracted by the exponential moving average of context encoder, i.e., target encoder, on the whole spectrogram. We find it beneficial to transfer random block masking into time-frequency aware masking in a curriculum manner, considering the complexity of highly correlated in local time and frequency in audio spectrograms. To enhance contextual semantic understanding and robustness, we fine-tune the encoder with a regularized masking on target datasets, instead of input dropping or zero. Empirically, when built with Vision Transformers structure, we find A-JEPA to be highly scalable and sets new state-of-the-art performance on multiple audio and speech classification tasks, outperforming other recent models that use externally supervised pre-training.

Image matching for both cross-view and cross-modality plays a critical role in multimodal perception. In practice, the modality gap caused by different imaging systems/styles poses great challenges to the matching task. Existing works try to extract invariant features for specific modalities and train on limited datasets, showing poor generalization. In this paper, we present MINIMA, a unified image matching framework for multiple cross-modal cases. Without pursuing fancy modules, our MINIMA aims to enhance universal performance from the perspective of data scaling up. For such purpose, we propose a simple yet effective data engine that can freely produce a large dataset containing multiple modalities, rich scenarios, and accurate matching labels. Specifically, we scale up the modalities from cheap but rich RGB-only matching data, by means of generative models. Under this setting, the matching labels and rich diversity of the RGB dataset are well inherited by the generated multimodal data. Benefiting from this, we construct MD-syn, a new comprehensive dataset that fills the data gap for general multimodal image matching. With MD-syn, we can directly train any advanced matching pipeline on randomly selected modality pairs to obtain cross-modal ability. Extensive experiments on in-domain and zero-shot matching tasks, including 19 cross-modal cases, demonstrate that our MINIMA can significantly outperform the baselines and even surpass modality-specific methods. The dataset and code are available at https://github.com/LSXI7/MINIMA .

Contrastive language-image pre-training (CLIP) has demonstrated remarkable zero-shot classification ability, namely image classification using novel text labels. Existing works have attempted to enhance CLIP by fine-tuning on downstream tasks, but these have inadvertently led to performance degradation on unseen classes, thus harming zero-shot generalization. This paper aims to address this challenge by leveraging readily available image-text pairs from an external dataset for cross-modal guidance during inference. To this end, we propose X-MoRe, a novel inference method comprising two key steps: (1) cross-modal retrieval and (2) modal-confidence-based ensemble. Given a query image, we harness the power of CLIP's cross-modal representations to retrieve relevant textual information from an external image-text pair dataset. Then, we assign higher weights to the more reliable modality between the original query image and retrieved text, contributing to the final prediction. X-MoRe demonstrates robust performance across a diverse set of tasks without the need for additional training, showcasing the effectiveness of utilizing cross-modal features to maximize CLIP's zero-shot ability.

Large Vision-Language Models (LVLMs) have achieved impressive performance, yet research has pointed out a serious issue with object hallucinations within these models. However, there is no clear conclusion as to which part of the model these hallucinations originate from. In this paper, we present an in-depth investigation into the object hallucination problem specifically within the CLIP model, which serves as the backbone for many state-of-the-art vision-language systems. We unveil that even in isolation, the CLIP model is prone to object hallucinations, suggesting that the hallucination problem is not solely due to the interaction between vision and language modalities. To address this, we propose a counterfactual data augmentation method by creating negative samples with a variety of hallucination issues. We demonstrate that our method can effectively mitigate object hallucinations for CLIP model, and we show the the enhanced model can be employed as a visual encoder, effectively alleviating the object hallucination issue in LVLMs.

In this paper, we tackle two challenges in multimodal learning for visual recognition: 1) when missing-modality occurs either during training or testing in real-world situations; and 2) when the computation resources are not available to finetune on heavy transformer models. To this end, we propose to utilize prompt learning and mitigate the above two challenges together. Specifically, our modality-missing-aware prompts can be plugged into multimodal transformers to handle general missing-modality cases, while only requiring less than 1% learnable parameters compared to training the entire model. We further explore the effect of different prompt configurations and analyze the robustness to missing modality. Extensive experiments are conducted to show the effectiveness of our prompt learning framework that improves the performance under various missing-modality cases, while alleviating the requirement of heavy model re-training. Code is available.

Vision-language pre-training like CLIP has shown promising performance on various downstream tasks such as zero-shot image classification and image-text retrieval. Most of the existing CLIP-alike works usually adopt relatively large image encoders like ResNet50 and ViT, while the lightweight counterparts are rarely discussed. In this paper, we propose a multi-level interaction paradigm for training lightweight CLIP models. Firstly, to mitigate the problem that some image-text pairs are not strictly one-to-one correspondence, we improve the conventional global instance-level alignment objective by softening the label of negative samples progressively. Secondly, a relaxed bipartite matching based token-level alignment objective is introduced for finer-grained alignment between image patches and textual words. Moreover, based on the observation that the accuracy of CLIP model does not increase correspondingly as the parameters of text encoder increase, an extra objective of masked language modeling (MLM) is leveraged for maximizing the potential of the shortened text encoder. In practice, an auxiliary fusion module injecting unmasked image embedding into masked text embedding at different network stages is proposed for enhancing the MLM. Extensive experiments show that without introducing additional computational cost during inference, the proposed method achieves a higher performance on multiple downstream tasks.

Multimodal models often experience a significant performance drop when one or more modalities are missing during inference. To address this challenge, we propose a simple yet effective approach that enhances robustness to missing modalities while maintaining strong performance when all modalities are available. Our method introduces cross-modal proxy tokens (CMPTs), which approximate the class token of a missing modality by attending only to the tokens of the available modality without requiring explicit modality generation or auxiliary networks. To efficiently learn these approximations with minimal computational overhead, we employ low-rank adapters in frozen unimodal encoders and jointly optimize an alignment loss with a task-specific loss. Extensive experiments on five multimodal datasets show that our method outperforms state-of-the-art baselines across various missing rates while achieving competitive results in complete-modality settings. Overall, our method offers a flexible and efficient solution for robust multimodal learning. The code and pretrained models will be released on GitHub.

Visible-infrared person re-identification (VIReID) primarily deals with matching identities across person images from different modalities. Due to the modality gap between visible and infrared images, cross-modality identity matching poses significant challenges. Recognizing that high-level semantics of pedestrian appearance, such as gender, shape, and clothing style, remain consistent across modalities, this paper intends to bridge the modality gap by infusing visual features with high-level semantics. Given the capability of CLIP to sense high-level semantic information corresponding to visual representations, we explore the application of CLIP within the domain of VIReID. Consequently, we propose a CLIP-Driven Semantic Discovery Network (CSDN) that consists of Modality-specific Prompt Learner, Semantic Information Integration (SII), and High-level Semantic Embedding (HSE). Specifically, considering the diversity stemming from modality discrepancies in language descriptions, we devise bimodal learnable text tokens to capture modality-private semantic information for visible and infrared images, respectively. Additionally, acknowledging the complementary nature of semantic details across different modalities, we integrate text features from the bimodal language descriptions to achieve comprehensive semantics. Finally, we establish a connection between the integrated text features and the visual features across modalities. This process embed rich high-level semantic information into visual representations, thereby promoting the modality invariance of visual representations. The effectiveness and superiority of our proposed CSDN over existing methods have been substantiated through experimental evaluations on multiple widely used benchmarks. The code will be released at https://github.com/nengdong96/CSDN.

Contrastive learning methods, such as CLIP, leverage naturally paired data-for example, images and their corresponding text captions-to learn general representations that transfer efficiently to downstream tasks. While such approaches are generally applied to two modalities, domains such as robotics, healthcare, and video need to support many types of data at once. We show that the pairwise application of CLIP fails to capture joint information between modalities, thereby limiting the quality of the learned representations. To address this issue, we present Symile, a simple contrastive learning approach that captures higher-order information between any number of modalities. Symile provides a flexible, architecture-agnostic objective for learning modality-specific representations. To develop Symile's objective, we derive a lower bound on total correlation, and show that Symile representations for any set of modalities form a sufficient statistic for predicting the remaining modalities. Symile outperforms pairwise CLIP, even with modalities missing in the data, on cross-modal classification and retrieval across several experiments including on an original multilingual dataset of 33M image, text and audio samples and a clinical dataset of chest X-rays, electrocardiograms, and laboratory measurements. All datasets and code used in this work are publicly available at https://github.com/rajesh-lab/symile.

Despite the success of Vision-Language Models (VLMs) like CLIP in aligning vision and language, their proficiency in detailed, fine-grained visual comprehension remains a key challenge. We present CLIP-IN, a novel framework that bolsters CLIP's fine-grained perception through two core innovations. Firstly, we leverage instruction-editing datasets, originally designed for image manipulation, as a unique source of hard negative image-text pairs. Coupled with a symmetric hard negative contrastive loss, this enables the model to effectively distinguish subtle visual-semantic differences. Secondly, CLIP-IN incorporates long descriptive captions, utilizing rotary positional encodings to capture rich semantic context often missed by standard CLIP. Our experiments demonstrate that CLIP-IN achieves substantial gains on the MMVP benchmark and various fine-grained visual recognition tasks, without compromising robust zero-shot performance on broader classification and retrieval tasks. Critically, integrating CLIP-IN's visual representations into Multimodal Large Language Models significantly reduces visual hallucinations and enhances reasoning abilities. This work underscores the considerable potential of synergizing targeted, instruction-based contrastive learning with comprehensive descriptive information to elevate the fine-grained understanding of VLMs.

Referring Remote Sensing Image Segmentation provides a flexible and fine-grained framework for remote sensing scene analysis via vision-language collaborative interpretation. Current approaches predominantly utilize a three-stage pipeline encompassing dual-modal encoding, cross-modal interaction, and pixel decoding. These methods demonstrate significant limitations in managing complex semantic relationships and achieving precise cross-modal alignment, largely due to their coupled processing mechanism that conflates target localization with boundary delineation. This architectural coupling amplifies error propagation under semantic ambiguity while restricting model generalizability and interpretability. To address these issues, we propose RSRefSeg 2, a decoupling paradigm that reformulates the conventional workflow into a collaborative dual-stage framework: coarse localization followed by fine segmentation. RSRefSeg 2 integrates CLIP's cross-modal alignment strength with SAM's segmentation generalizability through strategic foundation model collaboration. Specifically, CLIP is employed as the dual-modal encoder to activate target features within its pre-aligned semantic space and generate localization prompts. To mitigate CLIP's misactivation challenges in multi-entity scenarios described by referring texts, a cascaded second-order prompter is devised, which enhances precision through implicit reasoning via decomposition of text embeddings into complementary semantic subspaces. These optimized semantic prompts subsequently direct the SAM to generate pixel-level refined masks, thereby completing the semantic transmission pipeline. Extensive experiments (RefSegRS, RRSIS-D, and RISBench) demonstrate that RSRefSeg 2 surpasses contemporary methods in segmentation accuracy (+~3% gIoU) and complex semantic interpretation. Code is available at: https://github.com/KyanChen/RSRefSeg2.

CLIP is one of the most important multimodal foundational models today. What powers CLIP's capabilities? The rich supervision signals provided by natural language, the carrier of human knowledge, shape a powerful cross-modal representation space. However, with the rapid advancements in large language models LLMs like GPT-4 and LLaMA, the boundaries of language comprehension and generation are continually being pushed. This raises an intriguing question: can the capabilities of LLMs be harnessed to further improve multimodal representation learning? The potential benefits of incorporating LLMs into CLIP are clear. LLMs' strong textual understanding can fundamentally improve CLIP's ability to handle image captions, drastically enhancing its ability to process long and complex texts, a well-known limitation of vanilla CLIP. Moreover, LLMs are trained on a vast corpus of text, possessing open-world knowledge. This allows them to expand on caption information during training, increasing the efficiency of the learning process. In this paper, we propose LLM2CLIP, a novel approach that embraces the power of LLMs to unlock CLIP's potential. By fine-tuning the LLM in the caption space with contrastive learning, we extract its textual capabilities into the output embeddings, significantly improving the output layer's textual discriminability. We then design an efficient training process where the fine-tuned LLM acts as a powerful teacher for CLIP's visual encoder. Thanks to the LLM's presence, we can now incorporate longer and more complex captions without being restricted by vanilla CLIP's text encoder's context window and ability limitations. Our experiments demonstrate that this approach brings substantial improvements in cross-modal tasks.

Recently, Multimodal Large Language Models (MLLMs) have demonstrated impressive performance on instruction-following tasks by integrating pretrained visual encoders with large language models (LLMs). However, existing approaches often struggle to ground fine-grained visual concepts in complex scenes. In this paper, we propose MoDA (Modulation Adapter), a lightweight yet effective module designed to refine pre-aligned visual features through instruction-guided modulation. Our approach follows the standard LLaVA training protocol, consisting of a two-stage process: (1) aligning image features to the LLMs input space via a frozen vision encoder and adapter layers, and (2) refining those features using the MoDA adapter during the instructional tuning stage. MoDA employs a Transformer-based cross-attention mechanism to generate a modulation mask over the aligned visual tokens, thereby emphasizing semantically relevant embedding dimensions based on the language instruction. The modulated features are then passed to the LLM for autoregressive language generation. Our experimental evaluation shows that MoDA improves visual grounding and generates more contextually appropriate responses, demonstrating its effectiveness as a general-purpose enhancement for image-based MLLMs.

Multimodal large language models (MLLMs) recently showed strong capacity in integrating data among multiple modalities, empowered by a generalizable attention architecture. Advanced methods predominantly focus on language-centric tuning while less exploring multimodal tokens mixed through attention, posing challenges in high-level tasks that require fine-grained cognition and emotion understanding. In this work, we identify the attention deficit disorder problem in multimodal learning, caused by inconsistent cross-modal attention and layer-by-layer decayed attention activation. To address this, we propose a novel attention mechanism, termed MOdular Duplex Attention (MODA), simultaneously conducting the inner-modal refinement and inter-modal interaction. MODA employs a correct-after-align strategy to effectively decouple modality alignment from cross-layer token mixing. In the alignment phase, tokens are mapped to duplex modality spaces based on the basis vectors, enabling the interaction between visual and language modality. Further, the correctness of attention scores is ensured through adaptive masked attention, which enhances the model's flexibility by allowing customizable masking patterns for different modalities. Extensive experiments on 21 benchmark datasets verify the effectiveness of MODA in perception, cognition, and emotion tasks. Source code and demo are available in https://zzcheng.top/MODA.

Vanilla image completion approaches exhibit sensitivity to large missing regions, attributed to the limited availability of reference information for plausible generation. To mitigate this, existing methods incorporate the extra cue as a guidance for image completion. Despite improvements, these approaches are often restricted to employing a single modality (e.g., segmentation or sketch maps), which lacks scalability in leveraging multi-modality for more plausible completion. In this paper, we propose a novel, simple yet effective method for Multi-modal Guided Image Completion, dubbed MaGIC, which not only supports a wide range of single modality as the guidance (e.g., text, canny edge, sketch, segmentation, depth, and pose), but also adapts to arbitrarily customized combination of these modalities (i.e., arbitrary multi-modality) for image completion. For building MaGIC, we first introduce a modality-specific conditional U-Net (MCU-Net) that injects single-modal signal into a U-Net denoiser for single-modal guided image completion. Then, we devise a consistent modality blending (CMB) method to leverage modality signals encoded in multiple learned MCU-Nets through gradient guidance in latent space. Our CMB is training-free, thereby avoids the cumbersome joint re-training of different modalities, which is the secret of MaGIC to achieve exceptional flexibility in accommodating new modalities for completion. Experiments show the superiority of MaGIC over state-of-the-art methods and its generalization to various completion tasks. Our project with code and models is available at yeates.github.io/MaGIC-Page/.

Prompt tuning and adapter tuning have shown great potential in transferring pre-trained vision-language models (VLMs) to various downstream tasks. In this work, we design a new type of tuning method, termed as regularized mask tuning, which masks the network parameters through a learnable selection. Inspired by neural pathways, we argue that the knowledge required by a downstream task already exists in the pre-trained weights but just gets concealed in the upstream pre-training stage. To bring the useful knowledge back into light, we first identify a set of parameters that are important to a given downstream task, then attach a binary mask to each parameter, and finally optimize these masks on the downstream data with the parameters frozen. When updating the mask, we introduce a novel gradient dropout strategy to regularize the parameter selection, in order to prevent the model from forgetting old knowledge and overfitting the downstream data. Experimental results on 11 datasets demonstrate the consistent superiority of our method over previous alternatives. It is noteworthy that we manage to deliver 18.73% performance improvement compared to the zero-shot CLIP via masking an average of only 2.56% parameters. Furthermore, our method is synergistic with most existing parameter-efficient tuning methods and can boost the performance on top of them. Project page can be found here (https://wuw2019.github.io/R-AMT/).

Multimodal Large Language Models (MLLMs) suffer from cross-modal hallucinations, where one modality inappropriately influences generation about another, leading to fabricated output. This exposes a more fundamental deficiency in modality-interaction control. To address this, we propose Modality-Adaptive Decoding (MAD), a training-free method that adaptively weights modality-specific decoding branches based on task requirements. MAD leverages the model's inherent ability to self-assess modality relevance by querying which modalities are needed for each task. The extracted modality probabilities are then used to adaptively weight contrastive decoding branches, enabling the model to focus on relevant information while suppressing cross-modal interference. Extensive experiments on CMM and AVHBench demonstrate that MAD significantly reduces cross-modal hallucinations across multiple audio-visual language models (7.8\% and 2.0\% improvements for VideoLLaMA2-AV, 8.7\% and 4.7\% improvements for Qwen2.5-Omni). Our approach demonstrates that explicit modality awareness through self-assessment is crucial for robust multimodal reasoning, offering a principled extension to existing contrastive decoding methods. Our code is available at https://github.com/top-yun/MAD{https://github.com/top-yun/MAD}

Contrastive Language-Image Pre-Training (CLIP) is highly instrumental in machine learning applications within a large variety of domains. We investigate the geometry of this embedding, which is still not well understood. We examine the raw unnormalized embedding and show that text and image reside on linearly separable ellipsoid shells, not centered at the origin. We explain the benefits of having this structure, allowing to better embed instances according to their uncertainty during contrastive training. Frequent concepts in the dataset yield more false negatives, inducing greater uncertainty. A new notion of conformity is introduced, which measures the average cosine similarity of an instance to any other instance within a representative data set. We show this measure can be accurately estimated by simply computing the cosine similarity to the modality mean vector. Furthermore, we find that CLIP's modality gap optimizes the matching of the conformity distributions of image and text.

This paper introduces a powerful encoder that transfers CLIP`s capabilities to event-based data, enhancing its utility and expanding its applicability across diverse domains. While large-scale datasets have significantly advanced image-based models, the scarcity of comprehensive event datasets has limited performance potential in event modality. To address this challenge, we adapt CLIP`s architecture to align event embeddings with image embeddings, supporting zero-shot learning and preserving text alignment while mitigating catastrophic forgetting. Our encoder achieves strong performance in object recognition, with competitive results in zero-shot and few-shot learning tasks. Notably, it generalizes effectively to events extracted from video data without requiring additional training, highlighting its versatility. Additionally, we integrate this encoder within a cross-modality framework that facilitates interaction across five modalities-Image, Event, Text, Sound, and Depth-expanding the possibilities for cross-modal applications. Overall, this work underscores the transformative potential of a robust event encoder, broadening the scope and utility of event-based data across various fields.

Prompt learning is effective for fine-tuning foundation models to improve their generalization across a variety of downstream tasks. However, the prompts that are independently optimized along a single modality path, may sacrifice the vision-language alignment of pre-trained models in return for improved performance on specific tasks and classes, leading to poorer generalization. In this paper, we first demonstrate that prompt tuning along only one single branch of CLIP (e.g., language or vision) is the reason why the misalignment occurs. Without proper regularization across the learnable parameters in different modalities, prompt learning violates the original pre-training constraints inherent in the two-tower architecture. To address such misalignment, we first propose feature shift, which is defined as the variation of embeddings after introducing the learned prompts, to serve as an explanatory tool. We dive into its relation with generalizability and thereafter propose RESTORE, a multi-modal prompt learning method that exerts explicit constraints on cross-modal consistency. To be more specific, to prevent feature misalignment, a feature shift consistency is introduced to synchronize inter-modal feature shifts by measuring and regularizing the magnitude of discrepancy during prompt tuning. In addition, we propose a "surgery" block to avoid short-cut hacking, where cross-modal misalignment can still be severe if the feature shift of each modality varies drastically at the same rate. It is implemented as feed-forward adapters upon both modalities to alleviate the misalignment problem. Extensive experiments on 15 datasets demonstrate that our method outperforms the state-of-the-art prompt tuning methods without compromising feature alignment.

This paper extends Cross Modal Generalization (CMG) to open-set environments by proposing the more challenging Open-set Cross Modal Generalization (OSCMG) task. This task evaluates multimodal unified representations in open-set conditions, addressing the limitations of prior closed-set cross-modal evaluations. OSCMG requires not only cross-modal knowledge transfer but also robust generalization to unseen classes within new modalities, a scenario frequently encountered in real-world applications. Existing multimodal unified representation work lacks consideration for open-set environments. To tackle this, we propose MICU, comprising two key components: Fine-Coarse Masked multimodal InfoNCE (FCMI) and Cross modal Unified Jigsaw Puzzles (CUJP). FCMI enhances multimodal alignment by applying contrastive learning at both holistic semantic and temporal levels, incorporating masking to enhance generalization. CUJP enhances feature diversity and model uncertainty by integrating modality-agnostic feature selection with self-supervised learning, thereby strengthening the model's ability to handle unknown categories in open-set tasks. Extensive experiments on CMG and the newly proposed OSCMG validate the effectiveness of our approach. The code is available at https://github.com/haihuangcode/CMG.

Expanding multimodal representations to novel modalities is constrained by reliance on large-scale paired datasets (e.g., text-image, text-audio, text-3D, text-molecule), which are costly and often infeasible in domains requiring expert annotation such as medical imaging and molecular analysis. We introduce TextME, the first text-only modality expansion framework, to the best of our knowledge, projecting diverse modalities into LLM embedding space as a unified anchor. Our approach exploits the geometric structure of pretrained contrastive encoders to enable zero-shot cross-modal transfer using only text descriptions, without paired supervision. We empirically validate that such consistent modality gaps exist across image, video, audio, 3D, X-ray, and molecular domains, demonstrating that text-only training can preserve substantial performance of pretrained encoders. We further show that our framework enables emergent cross-modal retrieval between modality pairs not explicitly aligned during training (e.g., audio-to-image, 3D-to-image). These results establish text-only training as a practical alternative to paired supervision for modality expansion.

Recent unified multi-modal encoders align a wide range of modalities into a shared representation space, enabling diverse cross-modal tasks. Despite their impressive capabilities, the robustness of these models under adversarial perturbations remains underexplored, which is a critical concern for safety-sensitive applications. In this work, we present the first comprehensive study of adversarial vulnerability in unified multi-modal encoders. We find that even mild adversarial perturbations lead to substantial performance drops across all modalities. Non-visual inputs, such as audio and point clouds, are especially fragile, while visual inputs like images and videos also degrade significantly. To address this, we propose an efficient adversarial calibration framework that improves robustness across modalities without modifying pretrained encoders or semantic centers, ensuring compatibility with existing foundation models. Our method introduces modality-specific projection heads trained solely on adversarial examples, while keeping the backbone and embeddings frozen. We explore three training objectives: fixed-center cross-entropy, clean-to-adversarial L2 alignment, and clean-adversarial InfoNCE, and we introduce a regularization strategy to ensure modality-consistent alignment under attack. Experiments on six modalities and three Bind-style models show that our method improves adversarial robustness by up to 47.3 percent at epsilon = 4/255, while preserving or even improving clean zero-shot and retrieval performance with less than 1 percent trainable parameters.

Audio-Visual Large Language Models (AV-LLMs) face prohibitive computational overhead from massive audio and video tokens. Token reduction, while extensively explored for video-only LLMs, is insufficient for the audio-visual domain, as these unimodal methods cannot leverage audio-visual cross-modal synergies. Furthermore, the distinct and dynamic information densities of audio and video render static budgets per modality suboptimal. How to perform token reduction on a joint audio-visual stream thus remains an unaddressed bottleneck. To fill this gap, we introduce EchoingPixels, a framework inspired by the coexistence and interaction of visuals and sound in real-world scenes. The core of our framework is the Cross-Modal Semantic Sieve (CS2), a module enabling early audio-visual interaction. Instead of compressing modalities independently, CS2 co-attends to the joint multimodal stream and reduces tokens from an entire combined pool of audio-visual tokens rather than using fixed budgets per modality. This single-pool approach allows it to adaptively allocate the token budget across both modalities and dynamically identify salient tokens in concert. To ensure this aggressive reduction preserves the vital temporal modeling capability, we co-design a Synchronization-Augmented RoPE (Sync-RoPE) to maintain critical temporal relationships for the sparsely selected tokens. Extensive experiments demonstrate that EchoingPixels achieves performance comparable to strong baselines using only 5-20% of the original tokens, with a 2-3x speedup and memory reduction.

Multimodal models have gained significant success in recent years. Standard multimodal approaches often assume unchanged modalities from training stage to inference stage. In practice, however, many scenarios fail to satisfy such assumptions with missing modalities during inference, leading to limitations on where multimodal models can be applied. While existing methods mitigate the problem through reconstructing the missing modalities, it increases unnecessary computational cost, which could be just as critical, especially for large, deployed systems. To solve the problem from both sides, we propose a novel guidance network that promotes knowledge sharing during training, taking advantage of the multimodal representations to train better single-modality models for inference. Real-life experiment in violence detection shows that our proposed framework trains single-modality models that significantly outperform its traditionally trained counterparts while maintaining the same inference cost.

This paper investigates the performance of multimodal pre-trained models in user profiling tasks based on visual-linguistic demographic data. These models are critical for adapting to the needs and preferences of human users in social robotics, thereby providing personalized responses and enhancing interaction quality. First, we introduce two datasets specifically curated to represent demographic characteristics derived from user facial images. Next, we evaluate the performance of a prominent contrastive multimodal pre-trained model, CLIP, on these datasets, both in its out-of-the-box state and after fine-tuning. Initial results indicate that CLIP performs suboptimal in matching images to demographic descriptions without fine-tuning. Although fine-tuning significantly enhances its predictive capacity, the model continues to exhibit limitations in effectively generalizing subtle demographic nuances. To address this, we propose adopting a masked image modeling strategy to improve generalization and better capture subtle demographic attributes. This approach offers a pathway for enhancing demographic sensitivity in multimodal user modeling tasks.

Contrastive Language-Image Pre-training (CLIP) has become a cornerstone in vision-language representation learning, powering diverse downstream tasks and serving as the default vision backbone in multimodal large language models (MLLMs). Despite its success, CLIP's dense and opaque latent representations pose significant interpretability challenges. A common assumption is that interpretability and performance are in tension: enforcing sparsity during training degrades accuracy, motivating recent post-hoc approaches such as Sparse Autoencoders (SAEs). However, these post-hoc approaches often suffer from degraded downstream performance and loss of CLIP's inherent multimodal capabilities, with most learned features remaining unimodal. We propose a simple yet effective approach that integrates sparsity directly into CLIP training, yielding representations that are both interpretable and performant. Compared to SAEs, our Sparse CLIP representations preserve strong downstream task performance, achieve superior interpretability, and retain multimodal capabilities. We show that multimodal sparse features enable straightforward semantic concept alignment and reveal training dynamics of how cross-modal knowledge emerges. Finally, as a proof of concept, we train a vision-language model on sparse CLIP representations that enables interpretable, vision-based steering capabilities. Our findings challenge conventional wisdom that interpretability requires sacrificing accuracy and demonstrate that interpretability and performance can be co-optimized, offering a promising design principle for future models.

Audio and visual signals typically occur simultaneously, and humans possess an innate ability to correlate and synchronize information from these two modalities. Recently, a challenging problem known as Audio-Visual Segmentation (AVS) has emerged, intending to produce segmentation maps for sounding objects within a scene. However, the methods proposed so far have not sufficiently integrated audio and visual information, and the computational costs have been extremely high. Additionally, the outputs of different stages have not been fully utilized. To facilitate this research, we introduce a novel Progressive Confident Masking Attention Network (PMCANet). It leverages attention mechanisms to uncover the intrinsic correlations between audio signals and visual frames. Furthermore, we design an efficient and effective cross-attention module to enhance semantic perception by selecting query tokens. This selection is determined through confidence-driven units based on the network's multi-stage predictive outputs. Experiments demonstrate that our network outperforms other AVS methods while requiring less computational resources. The code is available at: https://github.com/PrettyPlate/PCMANet.

Recent video masked autoencoder (MAE) works have designed improved masking algorithms focused on saliency. These works leverage visual cues such as motion to mask the most salient regions. However, the robustness of such visual cues depends on how often input videos match underlying assumptions. On the other hand, natural language description is an information dense representation of video that implicitly captures saliency without requiring modality-specific assumptions, and has not been explored yet for video MAE. To this end, we introduce a novel text-guided masking algorithm (TGM) that masks the video regions with highest correspondence to paired captions. Without leveraging any explicit visual cues for saliency, our TGM is competitive with state-of-the-art masking algorithms such as motion-guided masking. To further benefit from the semantics of natural language for masked reconstruction, we next introduce a unified framework for joint MAE and masked video-text contrastive learning. We show that across existing masking algorithms, unifying MAE and masked video-text contrastive learning improves downstream performance compared to pure MAE on a variety of video recognition tasks, especially for linear probe. Within this unified framework, our TGM achieves the best relative performance on five action recognition and one egocentric datasets, highlighting the complementary nature of natural language for masked video modeling.

Vision-language models (VLMs) like CLIP have shown impressive zero-shot and few-shot learning capabilities across diverse applications. However, adapting these models to new fine-grained domains remains difficult due to reliance on prompt engineering and the high cost of full model fine-tuning. Existing adaptation approaches rely on augmented components, such as prompt tokens and adapter modules, which could limit adaptation quality, destabilize the model, and compromise the rich knowledge learned during pretraining. In this work, we present CLIP-SVD, a novel multi-modal and parameter-efficient adaptation technique that leverages Singular Value Decomposition (SVD) to modify the internal parameter space of CLIP without injecting additional modules. Specifically, we fine-tune only the singular values of the CLIP parameter matrices to rescale the basis vectors for domain adaptation while retaining the pretrained model. This design enables enhanced adaptation performance using only 0.04\% of the model's total parameters and better preservation of its generalization ability. CLIP-SVD achieves state-of-the-art classification results on 11 natural and 10 biomedical datasets, outperforming previous methods in both accuracy and generalization under few-shot settings. Additionally, we leverage a natural language-based approach to analyze the effectiveness and dynamics of the CLIP adaptation to allow interpretability of CLIP-SVD. The code is publicly available at https://github.com/HealthX-Lab/CLIP-SVD.

State Space Models (SSMs) have emerged as efficient alternatives to Transformers for sequential modeling, but their inability to leverage modality-specific features limits their performance in multi-modal pretraining. Here, we propose Mixture-of-Mamba, a novel SSM architecture that introduces modality-aware sparsity through modality-specific parameterization of the Mamba block. Building on Mixture-of-Transformers (W. Liang et al. arXiv:2411.04996; 2024), we extend the benefits of modality-aware sparsity to SSMs while preserving their computational efficiency. We evaluate Mixture-of-Mamba across three multi-modal pretraining settings: Transfusion (interleaved text and continuous image tokens with diffusion loss), Chameleon (interleaved text and discrete image tokens), and an extended three-modality framework incorporating speech. Mixture-of-Mamba consistently reaches the same loss values at earlier training steps with significantly reduced computational costs. In the Transfusion setting, Mixture-of-Mamba achieves equivalent image loss using only 34.76% of the training FLOPs at the 1.4B scale. In the Chameleon setting, Mixture-of-Mamba reaches similar image loss with just 42.50% of the FLOPs at the 1.4B scale, and similar text loss with just 65.40% of the FLOPs. In the three-modality setting, MoM matches speech loss at 24.80% of the FLOPs at the 1.4B scale. Our ablation study highlights the synergistic effects of decoupling projection components, where joint decoupling yields greater gains than individual modifications. These results establish modality-aware sparsity as a versatile and effective design principle, extending its impact from Transformers to SSMs and setting new benchmarks in multi-modal pretraining. Our code can be accessed at https://github.com/Weixin-Liang/Mixture-of-Mamba

Vision-Language Models (VLMs), such as CLIP, have demonstrated impressive zero-shot transfer capabilities in image-level visual perception. However, these models have shown limited performance in instance-level tasks that demand precise localization and recognition. Previous works have suggested that incorporating visual prompts, such as colorful boxes or circles, can improve the ability of models to recognize objects of interest. Nonetheless, compared to language prompting, visual prompting designs are rarely explored. Existing approaches, which employ coarse visual cues such as colorful boxes or circles, often result in sub-optimal performance due to the inclusion of irrelevant and noisy pixels. In this paper, we carefully study the visual prompting designs by exploring more fine-grained markings, such as segmentation masks and their variations. In addition, we introduce a new zero-shot framework that leverages pixel-level annotations acquired from a generalist segmentation model for fine-grained visual prompting. Consequently, our investigation reveals that a straightforward application of blur outside the target mask, referred to as the Blur Reverse Mask, exhibits exceptional effectiveness. This proposed prompting strategy leverages the precise mask annotations to reduce focus on weakly related regions while retaining spatial coherence between the target and the surrounding background. Our Fine-Grained Visual Prompting (FGVP) demonstrates superior performance in zero-shot comprehension of referring expressions on the RefCOCO, RefCOCO+, and RefCOCOg benchmarks. It outperforms prior methods by an average margin of 3.0% to 4.6%, with a maximum improvement of 12.5% on the RefCOCO+ testA subset. Code is available at https://github.com/ylingfeng/FGVP.

Models leveraging both visual and textual data such as Contrastive Language-Image Pre-training (CLIP), are the backbone of many recent advances in artificial intelligence. In this work, we show that despite their versatility, such models are vulnerable to what we refer to as fooling master images. Fooling master images are capable of maximizing the confidence score of a CLIP model for a significant number of widely varying prompts, while being either unrecognizable or unrelated to the attacked prompts for humans. The existence of such images is problematic as it could be used by bad actors to maliciously interfere with CLIP-trained image retrieval models in production with comparably small effort as a single image can attack many different prompts. We demonstrate how fooling master images for CLIP (CLIPMasterPrints) can be mined using stochastic gradient descent, projected gradient descent, or blackbox optimization. Contrary to many common adversarial attacks, the blackbox optimization approach allows us to mine CLIPMasterPrints even when the weights of the model are not accessible. We investigate the properties of the mined images, and find that images trained on a small number of image captions generalize to a much larger number of semantically related captions. We evaluate possible mitigation strategies, where we increase the robustness of the model and introduce an approach to automatically detect CLIPMasterPrints to sanitize the input of vulnerable models. Finally, we find that vulnerability to CLIPMasterPrints is related to a modality gap in contrastive pre-trained multi-modal networks. Code available at https://github.com/matfrei/CLIPMasterPrints.

Recent open-vocabulary segmentation methods adopt mask generators to predict segmentation masks and leverage pre-trained vision-language models, e.g., CLIP, to classify these masks via mask pooling. Although these approaches show promising results, it is counterintuitive that accurate masks often fail to yield accurate classification results through pooling CLIP image embeddings within the mask regions. In this paper, we reveal the performance limitations of mask pooling and introduce Mask-Adapter, a simple yet effective method to address these challenges in open-vocabulary segmentation. Compared to directly using proposal masks, our proposed Mask-Adapter extracts semantic activation maps from proposal masks, providing richer contextual information and ensuring alignment between masks and CLIP. Additionally, we propose a mask consistency loss that encourages proposal masks with similar IoUs to obtain similar CLIP embeddings to enhance models' robustness to varying predicted masks. Mask-Adapter integrates seamlessly into open-vocabulary segmentation methods based on mask pooling in a plug-and-play manner, delivering more accurate classification results. Extensive experiments across several zero-shot benchmarks demonstrate significant performance gains for the proposed Mask-Adapter on several well-established methods. Notably, Mask-Adapter also extends effectively to SAM and achieves impressive results on several open-vocabulary segmentation datasets. Code and models are available at https://github.com/hustvl/MaskAdapter.

Is vision good enough for language? Recent advancements in multimodal models primarily stem from the powerful reasoning abilities of large language models (LLMs). However, the visual component typically depends only on the instance-level contrastive language-image pre-training (CLIP). Our research reveals that the visual capabilities in recent multimodal LLMs (MLLMs) still exhibit systematic shortcomings. To understand the roots of these errors, we explore the gap between the visual embedding space of CLIP and vision-only self-supervised learning. We identify ''CLIP-blind pairs'' - images that CLIP perceives as similar despite their clear visual differences. With these pairs, we construct the Multimodal Visual Patterns (MMVP) benchmark. MMVP exposes areas where state-of-the-art systems, including GPT-4V, struggle with straightforward questions across nine basic visual patterns, often providing incorrect answers and hallucinated explanations. We further evaluate various CLIP-based vision-and-language models and found a notable correlation between visual patterns that challenge CLIP models and those problematic for multimodal LLMs. As an initial effort to address these issues, we propose a Mixture of Features (MoF) approach, demonstrating that integrating vision self-supervised learning features with MLLMs can significantly enhance their visual grounding capabilities. Together, our research suggests visual representation learning remains an open challenge, and accurate visual grounding is crucial for future successful multimodal systems.

Self-supervised representation learning follows a paradigm of withholding some part of the data and tasking the network to predict it from the remaining part. Among many techniques, data augmentation lies at the core for creating the information gap. Towards this end, masking has emerged as a generic and powerful tool where content is withheld along the sequential dimension, e.g., spatial in images, temporal in audio, and syntactic in language. In this paper, we explore the orthogonal channel dimension for generic data augmentation by exploiting precision redundancy. The data for each channel is quantized through a non-uniform quantizer, with the quantized value sampled randomly within randomly sampled quantization bins. From another perspective, quantization is analogous to channel-wise masking, as it removes the information within each bin, but preserves the information across bins. Our approach significantly surpasses existing generic data augmentation methods, while showing on par performance against modality-specific augmentations. We comprehensively evaluate our approach on vision, audio, 3D point clouds, as well as the DABS benchmark which is comprised of various data modalities. The code is available at https: //github.com/microsoft/random_quantize.

The ability to quickly learn a new task with minimal instruction - known as few-shot learning - is a central aspect of intelligent agents. Classical few-shot benchmarks make use of few-shot samples from a single modality, but such samples may not be sufficient to characterize an entire concept class. In contrast, humans use cross-modal information to learn new concepts efficiently. In this work, we demonstrate that one can indeed build a better {bf visual} dog classifier by {bf read}ing about dogs and {bf listen}ing to them bark. To do so, we exploit the fact that recent multimodal foundation models such as CLIP are inherently cross-modal, mapping different modalities to the same representation space. Specifically, we propose a simple cross-modal adaptation approach that learns from few-shot examples spanning different modalities. By repurposing class names as additional one-shot training samples, we achieve SOTA results with an embarrassingly simple linear classifier for vision-language adaptation. Furthermore, we show that our approach can benefit existing methods such as prefix tuning, adapters, and classifier ensembling. Finally, to explore other modalities beyond vision and language, we construct the first (to our knowledge) audiovisual few-shot benchmark and use cross-modal training to improve the performance of both image and audio classification.

Image modality is not perfect as it often fails in certain conditions, e.g., night and fast motion. This significantly limits the robustness and versatility of existing multi-modal (i.e., Image+X) semantic segmentation methods when confronting modality absence or failure, as often occurred in real-world applications. Inspired by the open-world learning capability of multi-modal vision-language models (MVLMs), we explore a new direction in learning the modality-agnostic representation via knowledge distillation (KD) from MVLMs. Intuitively, we propose Any2Seg, a novel framework that can achieve robust segmentation from any combination of modalities in any visual conditions. Specifically, we first introduce a novel language-guided semantic correlation distillation (LSCD) module to transfer both inter-modal and intra-modal semantic knowledge in the embedding space from MVLMs, e.g., LanguageBind. This enables us to minimize the modality gap and alleviate semantic ambiguity to combine any modalities in any visual conditions. Then, we introduce a modality-agnostic feature fusion (MFF) module that reweights the multi-modal features based on the inter-modal correlation and selects the fine-grained feature. This way, our Any2Seg finally yields an optimal modality-agnostic representation. Extensive experiments on two benchmarks with four modalities demonstrate that Any2Seg achieves the state-of-the-art under the multi-modal setting (+3.54 mIoU) and excels in the challenging modality-incomplete setting(+19.79 mIoU).

We tackle the challenge of open-vocabulary segmentation, where we need to identify objects from a wide range of categories in different environments, using text prompts as our input. To overcome this challenge, existing methods often use multi-modal models like CLIP, which combine image and text features in a shared embedding space to bridge the gap between limited and extensive vocabulary recognition, resulting in a two-stage approach: In the first stage, a mask generator takes an input image to generate mask proposals, and the in the second stage the target mask is picked based on the query. However, the expected target mask may not exist in the generated mask proposals, which leads to an unexpected output mask. In our work, we propose a novel approach named Prompt-guided Mask Proposal (PMP) where the mask generator takes the input text prompts and generates masks guided by these prompts. Compared with mask proposals generated without input prompts, masks generated by PMP are better aligned with the input prompts. To realize PMP, we designed a cross-attention mechanism between text tokens and query tokens which is capable of generating prompt-guided mask proposals after each decoding. We combined our PMP with several existing works employing a query-based segmentation backbone and the experiments on five benchmark datasets demonstrate the effectiveness of this approach, showcasing significant improvements over the current two-stage models (1% ~ 3% absolute performance gain in terms of mIOU). The steady improvement in performance across these benchmarks indicates the effective generalization of our proposed lightweight prompt-aware method.

Autonomous soundscape augmentation systems typically use trained models to pick optimal maskers to effect a desired perceptual change. While acoustic information is paramount to such systems, contextual information, including participant demographics and the visual environment, also influences acoustic perception. Hence, we propose modular modifications to an existing attention-based deep neural network, to allow early, mid-level, and late feature fusion of participant-linked, visual, and acoustic features. Ablation studies on module configurations and corresponding fusion methods using the ARAUS dataset show that contextual features improve the model performance in a statistically significant manner on the normalized ISO Pleasantness, to a mean squared error of 0.1194pm0.0012 for the best-performing all-modality model, against 0.1217pm0.0009 for the audio-only model. Soundscape augmentation systems can thereby leverage multimodal inputs for improved performance. We also investigate the impact of individual participant-linked factors using trained models to illustrate improvements in model explainability.

Recent advances in multimodal large language models (MLLMs) have enabled image-based question-answering capabilities. However, a key limitation is the use of CLIP as the visual encoder; while it can capture coarse global information, it often can miss fine-grained details that are relevant to the input query. To address these shortcomings, this work studies whether pre-trained text-to-image diffusion models can serve as instruction-aware visual encoders. Through an analysis of their internal representations, we find diffusion features are both rich in semantics and can encode strong image-text alignment. Moreover, we find that we can leverage text conditioning to focus the model on regions relevant to the input question. We then investigate how to align these features with large language models and uncover a leakage phenomenon, where the LLM can inadvertently recover information from the original diffusion prompt. We analyze the causes of this leakage and propose a mitigation strategy. Based on these insights, we explore a simple fusion strategy that utilizes both CLIP and conditional diffusion features. We evaluate our approach on both general VQA and specialized MLLM benchmarks, demonstrating the promise of diffusion models for visual understanding, particularly in vision-centric tasks that require spatial and compositional reasoning. Our project page can be found https://vatsalag99.github.io/mustafar/.

Prompt tuning (PT), as an emerging resource-efficient fine-tuning paradigm, has showcased remarkable effectiveness in improving the task-specific transferability of vision-language models. This paper delves into a previously overlooked information asymmetry issue in PT, where the visual modality mostly conveys more context than the object-oriented textual modality. Correspondingly, coarsely aligning these two modalities could result in the biased attention, driving the model to merely focus on the context area. To address this, we propose DAPT, an effective PT framework based on an intuitive decouple-before-align concept. First, we propose to explicitly decouple the visual modality into the foreground and background representation via exploiting coarse-and-fine visual segmenting cues, and then both of these decoupled patterns are aligned with the original foreground texts and the hand-crafted background classes, thereby symmetrically strengthening the modal alignment. To further enhance the visual concentration, we propose a visual pull-push regularization tailored for the foreground-background patterns, directing the original visual representation towards unbiased attention on the region-of-interest object. We demonstrate the power of architecture-free DAPT through few-shot learning, base-to-novel generalization, and data-efficient learning, all of which yield superior performance across prevailing benchmarks. Our code will be released at https://github.com/Ferenas/DAPT.

Multimodal learning with incomplete modality is practical and challenging. Recently, researchers have focused on enhancing the robustness of pre-trained MultiModal Transformers (MMTs) under missing modality conditions by applying learnable prompts. However, these prompt-based methods face several limitations: (1) incomplete modalities provide restricted modal cues for task-specific inference, (2) dummy imputation for missing content causes information loss and introduces noise, and (3) static prompts are instance-agnostic, offering limited knowledge for instances with various missing conditions. To address these issues, we propose RAGPT, a novel Retrieval-AuGmented dynamic Prompt Tuning framework. RAGPT comprises three modules: (I) the multi-channel retriever, which identifies similar instances through a within-modality retrieval strategy, (II) the missing modality generator, which recovers missing information using retrieved contexts, and (III) the context-aware prompter, which captures contextual knowledge from relevant instances and generates dynamic prompts to largely enhance the MMT's robustness. Extensive experiments conducted on three real-world datasets show that RAGPT consistently outperforms all competitive baselines in handling incomplete modality problems. The code of our work and prompt-based baselines is available at https://github.com/Jian-Lang/RAGPT.

The zero-shot performance of object detectors degrades when tested on different modalities, such as infrared and depth. While recent work has explored image translation techniques to adapt detectors to new modalities, these methods are limited to a single modality and apply only to traditional detectors. Recently, vision-language detectors, such as YOLO-World and Grounding DINO, have shown promising zero-shot capabilities, however, they have not yet been adapted for other visual modalities. Traditional fine-tuning approaches compromise the zero-shot capabilities of the detectors. The visual prompt strategies commonly used for classification with vision-language models apply the same linear prompt translation to each image, making them less effective. To address these limitations, we propose ModPrompt, a visual prompt strategy to adapt vision-language detectors to new modalities without degrading zero-shot performance. In particular, an encoder-decoder visual prompt strategy is proposed, further enhanced by the integration of inference-friendly modality prompt decoupled residual, facilitating a more robust adaptation. Empirical benchmarking results show our method for modality adaptation on two vision-language detectors, YOLO-World and Grounding DINO, and on challenging infrared (LLVIP, FLIR) and depth (NYUv2) datasets, achieving performance comparable to full fine-tuning while preserving the model's zero-shot capability. Code available at: https://github.com/heitorrapela/ModPrompt.

Recent text-to-video diffusion models can generate compelling video sequences, yet they remain silent -- missing the semantic, emotional, and atmospheric cues that audio provides. We introduce LTX-2, an open-source foundational model capable of generating high-quality, temporally synchronized audiovisual content in a unified manner. LTX-2 consists of an asymmetric dual-stream transformer with a 14B-parameter video stream and a 5B-parameter audio stream, coupled through bidirectional audio-video cross-attention layers with temporal positional embeddings and cross-modality AdaLN for shared timestep conditioning. This architecture enables efficient training and inference of a unified audiovisual model while allocating more capacity for video generation than audio generation. We employ a multilingual text encoder for broader prompt understanding and introduce a modality-aware classifier-free guidance (modality-CFG) mechanism for improved audiovisual alignment and controllability. Beyond generating speech, LTX-2 produces rich, coherent audio tracks that follow the characters, environment, style, and emotion of each scene -- complete with natural background and foley elements. In our evaluations, the model achieves state-of-the-art audiovisual quality and prompt adherence among open-source systems, while delivering results comparable to proprietary models at a fraction of their computational cost and inference time. All model weights and code are publicly released.

We present XKD, a novel self-supervised framework to learn meaningful representations from unlabelled videos. XKD is trained with two pseudo objectives. First, masked data reconstruction is performed to learn modality-specific representations from audio and visual streams. Next, self-supervised cross-modal knowledge distillation is performed between the two modalities through a teacher-student setup to learn complementary information. We introduce a novel domain alignment strategy to tackle domain discrepancy between audio and visual modalities enabling effective cross-modal knowledge distillation. Additionally, to develop a general-purpose network capable of handling both audio and visual streams, modality-agnostic variants of XKD are introduced, which use the same pretrained backbone for different audio and visual tasks. Our proposed cross-modal knowledge distillation improves video action classification by 8% to 14% on UCF101, HMDB51, and Kinetics400. Additionally, XKD improves multimodal action classification by 5.5% on Kinetics-Sound. XKD shows state-of-the-art performance in sound classification on ESC50, achieving top-1 accuracy of 96.5%.

Contrastive audio-language pretraining yields powerful joint representations, yet a persistent audio-text modality gap limits the benefits of coupling multimodal encoders with large language models (LLMs). We present Diffusion-Link, a diffusion-based modality-bridging module that generatively maps audio embeddings into the text-embedding distribution. The module is trained at the output embedding from the frozen multimodal encoder and implemented as a lightweight network with three residual MLP blocks. To assess the effect of Diffusion-Link on multimodal encoder-LLM coupling, we evaluate on Automatic Audio Captioning (AAC); to our knowledge, this is the first application of diffusion-based modality bridging to AAC. We report two results. (1) Modality-gap analysis: on similarity and geometric criteria, Diffusion-Link reduces the modality gap the most among prior diffusion-based methods and shows a collective migration of audio embeddings toward the text distribution. (2) Downstream AAC: attaching Diffusion-Link to the same multimodal LLM baseline achieves state-of-the-art on AudioCaps in both zero-shot and fully supervised captioning without external knowledge, with relative gains up to 52.5% and 7.5%, respectively. These findings show that closing the modality gap is pivotal for effective coupling between multimodal encoders and LLMs, and diffusion-based modality bridging offers a promising direction beyond knowledge-retrieval-centric designs. Code will be released upon acceptance https://github.com/DevKiHyun/Diffusion-Link

Multi-modal (vision-language) models, such as CLIP, are replacing traditional supervised pre-training models (e.g., ImageNet-based pre-training) as the new generation of visual foundation models. These models with robust and aligned semantic representations learned from billions of internet image-text pairs and can be applied to various downstream tasks in a zero-shot manner. However, in some fine-grained domains like medical imaging and remote sensing, the performance of multi-modal foundation models often leaves much to be desired. Consequently, many researchers have begun to explore few-shot adaptation methods for these models, gradually deriving three main technical approaches: 1) prompt-based methods, 2) adapter-based methods, and 3) external knowledge-based methods. Nevertheless, this rapidly developing field has produced numerous results without a comprehensive survey to systematically organize the research progress. Therefore, in this survey, we introduce and analyze the research advancements in few-shot adaptation methods for multi-modal models, summarizing commonly used datasets and experimental setups, and comparing the results of different methods. In addition, due to the lack of reliable theoretical support for existing methods, we derive the few-shot adaptation generalization error bound for multi-modal models. The theorem reveals that the generalization error of multi-modal foundation models is constrained by three factors: domain gap, model capacity, and sample size. Based on this, we propose three possible solutions from the following aspects: 1) adaptive domain generalization, 2) adaptive model selection, and 3) adaptive knowledge utilization.

A jump cut offers an abrupt, sometimes unwanted change in the viewing experience. We present a novel framework for smoothing these jump cuts, in the context of talking head videos. We leverage the appearance of the subject from the other source frames in the video, fusing it with a mid-level representation driven by DensePose keypoints and face landmarks. To achieve motion, we interpolate the keypoints and landmarks between the end frames around the cut. We then use an image translation network from the keypoints and source frames, to synthesize pixels. Because keypoints can contain errors, we propose a cross-modal attention scheme to select and pick the most appropriate source amongst multiple options for each key point. By leveraging this mid-level representation, our method can achieve stronger results than a strong video interpolation baseline. We demonstrate our method on various jump cuts in the talking head videos, such as cutting filler words, pauses, and even random cuts. Our experiments show that we can achieve seamless transitions, even in the challenging cases where the talking head rotates or moves drastically in the jump cut.

CLIP models learn transferable multi-modal features via image-text contrastive learning on internet-scale data. They are widely used in zero-shot classification, multi-modal retrieval, text-to-image diffusion, and as image encoders in large vision-language models. However, CLIP's pretraining is dominated by images paired with short captions, biasing the model toward encoding simple descriptions of salient objects and leading to coarse alignment on complex scenes and dense descriptions. While recent work mitigates this by fine-tuning on small-scale long-caption datasets, we identify an important common bias: both human- and LLM-generated long captions typically begin with a one-sentence summary followed by a detailed description. We show that this acts as a shortcut during training, concentrating attention on the opening sentence and early tokens and weakening alignment over the rest of the caption. To resolve this, we introduce DeBias-CLIP, which removes the summary sentence during training and applies sentence sub-sampling and text token padding to distribute supervision across all token positions. DeBias-CLIP achieves state-of-the-art long-text retrieval, improves short-text retrieval, and is less sensitive to sentence order permutations. It is a drop-in replacement for Long-CLIP with no additional trainable parameters.

Contrastive Language-Image Pre-training (CLIP), which excels at abstracting open-world representations across domains and modalities, has become a foundation for a variety of vision and multimodal tasks. However, recent studies reveal that CLIP has severe visual shortcomings, such as which can hardly distinguish orientation, quantity, color, structure, etc. These visual shortcomings also limit the perception capabilities of multimodal large language models (MLLMs) built on CLIP. The main reason could be that the image-text pairs used to train CLIP are inherently biased, due to the lack of the distinctiveness of the text and the diversity of images. In this work, we present a simple post-training approach for CLIP models, which largely overcomes its visual shortcomings via a self-supervised diffusion process. We introduce DIVA, which uses the DIffusion model as a Visual Assistant for CLIP. Specifically, DIVA leverages generative feedback from text-to-image diffusion models to optimize CLIP representations, with only images (without corresponding text). We demonstrate that DIVA improves CLIP's performance on the challenging MMVP-VLM benchmark which assesses fine-grained visual abilities to a large extent (e.g., 3-7%), and enhances the performance of MLLMs and vision models on multimodal understanding and segmentation tasks. Extensive evaluation on 29 image classification and retrieval benchmarks confirms that our framework preserves CLIP's strong zero-shot capabilities. The code will be available at https://github.com/baaivision/DIVA.

Deepfake technology has rapidly advanced and poses significant threats to information integrity and trust in online multimedia. While significant progress has been made in detecting deepfakes, the simultaneous manipulation of audio and visual modalities, sometimes at small parts or in subtle ways, presents highly challenging detection scenarios. To address these challenges, we present DiMoDif, an audio-visual deepfake detection framework that leverages the inter-modality differences in machine perception of speech, based on the assumption that in real samples -- in contrast to deepfakes -- visual and audio signals coincide in terms of information. DiMoDif leverages features from deep networks that specialize in visual and audio speech recognition to spot frame-level cross-modal incongruities, and in that way to temporally localize the deepfake forgery. To this end, we devise a hierarchical cross-modal fusion network, integrating adaptive temporal alignment modules and a learned discrepancy mapping layer to explicitly model the subtle differences between visual and audio representations. Then, the detection model is optimized through a composite loss function accounting for frame-level detections and fake intervals localization. DiMoDif outperforms the state-of-the-art on the Deepfake Detection task by 30.5 AUC on the highly challenging AV-Deepfake1M, while it performs exceptionally on FakeAVCeleb and LAV-DF. On the Temporal Forgery Localization task, it outperforms the state-of-the-art by 47.88 AP@0.75 on AV-Deepfake1M, and performs on-par on LAV-DF. Code available at https://github.com/mever-team/dimodif.

Massive multi-modality datasets play a significant role in facilitating the success of large video-language models. However, current video-language datasets primarily provide text descriptions for visual frames, considering audio to be weakly related information. They usually overlook exploring the potential of inherent audio-visual correlation, leading to monotonous annotation within each modality instead of comprehensive and precise descriptions. Such ignorance results in the difficulty of multiple cross-modality studies. To fulfill this gap, we present MMTrail, a large-scale multi-modality video-language dataset incorporating more than 20M trailer clips with visual captions, and 2M high-quality clips with multimodal captions. Trailers preview full-length video works and integrate context, visual frames, and background music. In particular, the trailer has two main advantages: (1) the topics are diverse, and the content characters are of various types, e.g., film, news, and gaming. (2) the corresponding background music is custom-designed, making it more coherent with the visual context. Upon these insights, we propose a systemic captioning framework, achieving various modality annotations with more than 27.1k hours of trailer videos. Here, to ensure the caption retains music perspective while preserving the authority of visual context, we leverage the advanced LLM to merge all annotations adaptively. In this fashion, our MMtrail dataset potentially paves the path for fine-grained large multimodal-language model training. In experiments, we provide evaluation metrics and benchmark results on our dataset, demonstrating the high quality of our annotation and its effectiveness for model training.

Large-scale multi-modal training with image-text pairs imparts strong generalization to CLIP model. Since training on a similar scale for videos is infeasible, recent approaches focus on the effective transfer of image-based CLIP to the video domain. In this pursuit, new parametric modules are added to learn temporal information and inter-frame relationships which require meticulous design efforts. Furthermore, when the resulting models are learned on videos, they tend to overfit on the given task distribution and lack in generalization aspect. This begs the following question: How to effectively transfer image-level CLIP representations to videos? In this work, we show that a simple Video Fine-tuned CLIP (ViFi-CLIP) baseline is generally sufficient to bridge the domain gap from images to videos. Our qualitative analysis illustrates that the frame-level processing from CLIP image-encoder followed by feature pooling and similarity matching with corresponding text embeddings helps in implicitly modeling the temporal cues within ViFi-CLIP. Such fine-tuning helps the model to focus on scene dynamics, moving objects and inter-object relationships. For low-data regimes where full fine-tuning is not viable, we propose a `bridge and prompt' approach that first uses fine-tuning to bridge the domain gap and then learns prompts on language and vision side to adapt CLIP representations. We extensively evaluate this simple yet strong baseline on zero-shot, base-to-novel generalization, few-shot and fully supervised settings across five video benchmarks. Our code is available at https://github.com/muzairkhattak/ViFi-CLIP.

We present Fast Language-Image Pre-training (FLIP), a simple and more efficient method for training CLIP. Our method randomly masks out and removes a large portion of image patches during training. Masking allows us to learn from more image-text pairs given the same wall-clock time and contrast more samples per iteration with similar memory footprint. It leads to a favorable trade-off between accuracy and training time. In our experiments on 400 million image-text pairs, FLIP improves both accuracy and speed over the no-masking baseline. On a large diversity of downstream tasks, FLIP dominantly outperforms the CLIP counterparts trained on the same data. Facilitated by the speedup, we explore the scaling behavior of increasing the model size, data size, or training length, and report encouraging results and comparisons. We hope that our work will foster future research on scaling vision-language learning.

Video-language pre-training is crucial for learning powerful multi-modal representation. However, it typically requires a massive amount of computation. In this paper, we develop SMAUG, an efficient pre-training framework for video-language models. The foundation component in SMAUG is masked autoencoders. Different from prior works which only mask textual inputs, our masking strategy considers both visual and textual modalities, providing a better cross-modal alignment and saving more pre-training costs. On top of that, we introduce a space-time token sparsification module, which leverages context information to further select only "important" spatial regions and temporal frames for pre-training. Coupling all these designs allows our method to enjoy both competitive performances on text-to-video retrieval and video question answering tasks, and much less pre-training costs by 1.9X or more. For example, our SMAUG only needs about 50 NVIDIA A6000 GPU hours for pre-training to attain competitive performances on these two video-language tasks across six popular benchmarks.

Audio-visual automatic speech recognition (AV-ASR) is an extension of ASR that incorporates visual cues, often from the movements of a speaker's mouth. Unlike works that simply focus on the lip motion, we investigate the contribution of entire visual frames (visual actions, objects, background etc.). This is particularly useful for unconstrained videos, where the speaker is not necessarily visible. To solve this task, we propose a new sequence-to-sequence AudioVisual ASR TrAnsformeR (AVATAR) which is trained end-to-end from spectrograms and full-frame RGB. To prevent the audio stream from dominating training, we propose different word-masking strategies, thereby encouraging our model to pay attention to the visual stream. We demonstrate the contribution of the visual modality on the How2 AV-ASR benchmark, especially in the presence of simulated noise, and show that our model outperforms all other prior work by a large margin. Finally, we also create a new, real-world test bed for AV-ASR called VisSpeech, which demonstrates the contribution of the visual modality under challenging audio conditions.

Multimodal Large Language Models excel in various tasks, yet often struggle with modality bias, where the model tends to rely heavily on a single modality and overlook critical information in other modalities, which leads to incorrect focus and generating irrelevant responses. In this paper, we propose using the paradigm of preference optimization to solve the modality bias problem, including RLAIFVBias, a debiased preference optimization dataset, and a Noise Aware Preference Optimization algorithm. Specifically, we first construct the dataset by introducing perturbations to reduce the informational content of certain modalities, compelling the model to rely on a specific modality when generating negative responses. To address the inevitable noise in automatically constructed data, we combine the noise robust Mean Absolute Error with the Binary Cross Entropy in Direct Preference Optimization by a negative Box Cox transformation, and dynamically adjust the algorithm noise robustness based on the evaluated noise levels in the data. Extensive experiments validate our approach, demonstrating not only its effectiveness in mitigating modality bias but also its significant role in minimizing hallucinations.

We introduce Audio-Visual Affordance Grounding (AV-AG), a new task that segments object interaction regions from action sounds. Unlike existing approaches that rely on textual instructions or demonstration videos, which often limited by ambiguity or occlusion, audio provides real-time, semantically rich, and visually independent cues for affordance grounding, enabling more intuitive understanding of interaction regions. To support this task, we construct the first AV-AG dataset, comprising a large collection of action sounds, object images, and pixel-level affordance annotations. The dataset also includes an unseen subset to evaluate zero-shot generalization. Furthermore, we propose AVAGFormer, a model equipped with a semantic-conditioned cross-modal mixer and a dual-head decoder that effectively fuses audio and visual signals for mask prediction. Experiments show that AVAGFormer achieves state-of-the-art performance on AV-AG, surpassing baselines from related tasks. Comprehensive analyses highlight the distinctions between AV-AG and AVS, the benefits of end-to-end modeling, and the contribution of each component. Code and dataset have been released on https://jscslld.github.io/AVAGFormer/.

Zero-shot, training-free, image-based text-to-video generation is an emerging area that aims to generate videos using existing image-based diffusion models. Current methods in this space require specific architectural changes to image generation models, which limit their adaptability and scalability. In contrast to such methods, we provide a model-agnostic approach. We use intersections in diffusion trajectories, working only with the latent values. We could not obtain localized frame-wise coherence and diversity using only the intersection of trajectories. Thus, we instead use a grid-based approach. An in-context trained LLM is used to generate coherent frame-wise prompts; another is used to identify differences between frames. Based on these, we obtain a CLIP-based attention mask that controls the timing of switching the prompts for each grid cell. Earlier switching results in higher variance, while later switching results in more coherence. Therefore, our approach can ensure appropriate control between coherence and variance for the frames. Our approach results in state-of-the-art performance while being more flexible when working with diverse image-generation models. The empirical analysis using quantitative metrics and user studies confirms our model's superior temporal consistency, visual fidelity and user satisfaction, thus providing a novel way to obtain training-free, image-based text-to-video generation.

Audio-visual segmentation aims to separate sounding objects from videos by predicting pixel-level masks based on audio signals. Existing methods primarily concentrate on closed-set scenarios and direct audio-visual alignment and fusion, which limits their capability to generalize to new, unseen situations. In this paper, we propose OpenAVS, a novel training-free language-based approach that, for the first time, effectively aligns audio and visual modalities using text as a proxy for open-vocabulary Audio-Visual Segmentation (AVS). Equipped with multimedia foundation models, OpenAVS directly infers masks through 1) audio-to-text prompt generation, 2) LLM-guided prompt translation, and 3) text-to-visual sounding object segmentation. The objective of OpenAVS is to establish a simple yet flexible architecture that relies on the most appropriate foundation models by fully leveraging their capabilities to enable more effective knowledge transfer to the downstream AVS task. Moreover, we present a model-agnostic framework OpenAVS-ST that enables the integration of OpenAVS with any advanced supervised AVS model via pseudo-label based self-training. This approach enhances performance by effectively utilizing large-scale unlabeled data when available. Comprehensive experiments on three benchmark datasets demonstrate the superior performance of OpenAVS. It surpasses existing unsupervised, zero-shot, and few-shot AVS methods by a significant margin, achieving absolute performance gains of approximately 9.4% and 10.9% in mIoU and F-score, respectively, in challenging scenarios.

Dynamic Facial Expression Recognition (DFER) is crucial for understanding human behavior. However, current methods exhibit limited performance mainly due to the scarcity of high-quality data, the insufficient utilization of facial dynamics, and the ambiguity of expression semantics, etc. To this end, we propose a novel framework, named Multi-modal Fine-grained CLIP for Dynamic Facial Expression Recognition with AdaptERs (FineCLIPER), incorporating the following novel designs: 1) To better distinguish between similar facial expressions, we extend the class labels to textual descriptions from both positive and negative aspects, and obtain supervision by calculating the cross-modal similarity based on the CLIP model; 2) Our FineCLIPER adopts a hierarchical manner to effectively mine useful cues from DFE videos. Specifically, besides directly embedding video frames as input (low semantic level), we propose to extract the face segmentation masks and landmarks based on each frame (middle semantic level) and utilize the Multi-modal Large Language Model (MLLM) to further generate detailed descriptions of facial changes across frames with designed prompts (high semantic level). Additionally, we also adopt Parameter-Efficient Fine-Tuning (PEFT) to enable efficient adaptation of large pre-trained models (i.e., CLIP) for this task. Our FineCLIPER achieves SOTA performance on the DFEW, FERV39k, and MAFW datasets in both supervised and zero-shot settings with few tunable parameters. Project Page: https://haroldchen19.github.io/FineCLIPER-Page/

Mixture-of-Experts (MoE) Multimodal large language models (MLLMs) excel at vision-language tasks, but they suffer from high computational inefficiency. To reduce inference overhead, expert skipping methods have been proposed to deactivate redundant experts based on the current input tokens. However, we find that applying these methods-originally designed for unimodal large language models (LLMs)-to MLLMs results in considerable performance degradation. This is primarily because such methods fail to account for the heterogeneous contributions of experts across MoE layers and modality-specific behaviors of tokens within these layers. Motivated by these findings, we propose MoDES, the first training-free framework that adaptively skips experts to enable efficient and accurate MoE MLLM inference. It incorporates a globally-modulated local gating (GMLG) mechanism that integrates global layer-wise importance into local routing probabilities to accurately estimate per-token expert importance. A dual-modality thresholding (DMT) method is then applied, which processes tokens from each modality separately, to derive the skipping schedule. To set the optimal thresholds, we introduce a frontier search algorithm that exploits monotonicity properties, cutting convergence time from several days to a few hours. Extensive experiments for 3 model series across 13 benchmarks demonstrate that MoDES far outperforms previous approaches. For instance, when skipping 88% experts for Qwen3-VL-MoE-30B-A3B-Instruct, the performance boost is up to 10.67% (97.33% vs. 86.66%). Furthermore, MoDES significantly enhances inference speed, improving the prefilling time by 2.16times and the decoding time by 1.26times.

CLIP models perform remarkably well on zero-shot classification and retrieval tasks. But recent studies have shown that learnt representations in CLIP are not well suited for dense prediction tasks like object detection, semantic segmentation or depth estimation. More recently, multi-stage training methods for CLIP models was introduced to mitigate the weak performance of CLIP on downstream tasks. In this work, we find that simply improving the quality of captions in image-text datasets improves the quality of CLIP's visual representations, resulting in significant improvement on downstream dense prediction vision tasks. In fact, we find that CLIP pretraining with good quality captions can surpass recent supervised, self-supervised and weakly supervised pretraining methods. We show that when CLIP model with ViT-B/16 as image encoder is trained on well aligned image-text pairs it obtains 12.1% higher mIoU and 11.5% lower RMSE on semantic segmentation and depth estimation tasks over recent state-of-the-art Masked Image Modeling (MIM) pretraining methods like Masked Autoencoder (MAE). We find that mobile architectures also benefit significantly from CLIP pretraining. A recent mobile vision architecture, MCi2, with CLIP pretraining obtains similar performance as Swin-L, pretrained on ImageNet-22k for semantic segmentation task while being 6.1times smaller. Moreover, we show that improving caption quality results in 10times data efficiency when finetuning for dense prediction tasks.

Recovering the masked speech frames is widely applied in speech representation learning. However, most of these models use random masking in the pre-training. In this work, we proposed two kinds of masking approaches: (1) speech-level masking, making the model to mask more speech segments than silence segments, (2) phoneme-level masking, forcing the model to mask the whole frames of the phoneme, instead of phoneme pieces. We pre-trained the model via these two approaches, and evaluated on two downstream tasks, phoneme classification and speaker recognition. The experiments demonstrated that the proposed masking approaches are beneficial to improve the performance of speech representation.

In this paper, we present IMAGEdit, a training-free framework for any number of video subject editing that manipulates the appearances of multiple designated subjects while preserving non-target regions, without finetuning or retraining. We achieve this by providing robust multimodal conditioning and precise mask sequences through a prompt-guided multimodal alignment module and a prior-based mask retargeting module. We first leverage large models' understanding and generation capabilities to produce multimodal information and mask motion sequences for multiple subjects across various types. Then, the obtained prior mask sequences are fed into a pretrained mask-driven video generation model to synthesize the edited video. With strong generalization capability, IMAGEdit remedies insufficient prompt-side multimodal conditioning and overcomes mask boundary entanglement in videos with any number of subjects, thereby significantly expanding the applicability of video editing. More importantly, IMAGEdit is compatible with any mask-driven video generation model, significantly improving overall performance. Extensive experiments on our newly constructed multi-subject benchmark MSVBench verify that IMAGEdit consistently surpasses state-of-the-art methods. Code, models, and datasets are publicly available at https://github.com/XWH-A/IMAGEdit.

Multi-modal foundation models like OpenFlamingo, LLaVA, and GPT-4 are increasingly used for various real-world tasks. Prior work has shown that these models are highly vulnerable to adversarial attacks on the vision modality. These attacks can be leveraged to spread fake information or defraud users, and thus pose a significant risk, which makes the robustness of large multi-modal foundation models a pressing problem. The CLIP model, or one of its variants, is used as a frozen vision encoder in many vision-language models (VLMs), e.g. LLaVA and OpenFlamingo. We propose an unsupervised adversarial fine-tuning scheme to obtain a robust CLIP vision encoder, which yields robustness on all vision down-stream tasks (VLMs, zero-shot classification) that rely on CLIP. In particular, we show that stealth-attacks on users of VLMs by a malicious third party providing manipulated images are no longer possible once one replaces the original CLIP model with our robust one. No retraining or fine-tuning of the VLM is required. The code and robust models are available at https://github.com/chs20/RobustVLM

Contrastive Language-Image Pre-training (CLIP) plays an essential role in extracting valuable content information from images across diverse tasks. It aligns textual and visual modalities to comprehend the entire image, including all the details, even those irrelevant to specific tasks. However, for a finer understanding and controlled editing of images, it becomes crucial to focus on specific regions of interest, which can be indicated as points, masks, or boxes by humans or perception models. To fulfill the requirements, we introduce Alpha-CLIP, an enhanced version of CLIP with an auxiliary alpha channel to suggest attentive regions and fine-tuned with constructed millions of RGBA region-text pairs. Alpha-CLIP not only preserves the visual recognition ability of CLIP but also enables precise control over the emphasis of image contents. It demonstrates effectiveness in various tasks, including but not limited to open-world recognition, multimodal large language models, and conditional 2D / 3D generation. It has a strong potential to serve as a versatile tool for image-related tasks.

End-to-end human animation with rich multi-modal conditions, e.g., text, image and audio has achieved remarkable advancements in recent years. However, most existing methods could only animate a single subject and inject conditions in a global manner, ignoring scenarios that multiple concepts could appears in the same video with rich human-human interactions and human-object interactions. Such global assumption prevents precise and per-identity control of multiple concepts including humans and objects, therefore hinders applications. In this work, we discard the single-entity assumption and introduce a novel framework that enforces strong, region-specific binding of conditions from modalities to each identity's spatiotemporal footprint. Given reference images of multiple concepts, our method could automatically infer layout information by leveraging a mask predictor to match appearance cues between the denoised video and each reference appearance. Furthermore, we inject local audio condition into its corresponding region to ensure layout-aligned modality matching in a iterative manner. This design enables the high-quality generation of controllable multi-concept human-centric videos. Empirical results and ablation studies validate the effectiveness of our explicit layout control for multi-modal conditions compared to implicit counterparts and other existing methods.

Within the domain of medical analysis, extensive research has explored the potential of mutual learning between Masked Autoencoders(MAEs) and multimodal data. However, the impact of MAEs on intermodality remains a key challenge. We introduce MedFLIP, a Fast Language-Image Pre-training method for Medical analysis. We explore MAEs for zero-shot learning with crossed domains, which enhances the model's ability to learn from limited data, a common scenario in medical diagnostics. We verify that masking an image does not affect inter-modal learning. Furthermore, we propose the SVD loss to enhance the representation learning for characteristics of medical images, aiming to improve classification accuracy by leveraging the structural intricacies of such data. Our theory posits that masking encourages semantic preservation, robust feature extraction, regularization, domain adaptation, and invariance learning. Lastly, we validate using language will improve the zero-shot performance for the medical image analysis. MedFLIP's scaling of the masking process marks an advancement in the field, offering a pathway to rapid and precise medical image analysis without the traditional computational bottlenecks. Through experiments and validation, MedFLIP demonstrates efficient performance improvements, helps for future research and application in medical diagnostics.

Video generation has made substantial strides with the emergence of deep generative models, especially diffusion-based approaches. However, video generation based on multiple reference subjects still faces significant challenges in maintaining multi-subject consistency and ensuring high generation quality. In this paper, we propose MAGREF, a unified framework for any-reference video generation that introduces masked guidance to enable coherent multi-subject video synthesis conditioned on diverse reference images and a textual prompt. Specifically, we propose (1) a region-aware dynamic masking mechanism that enables a single model to flexibly handle various subject inference, including humans, objects, and backgrounds, without architectural changes, and (2) a pixel-wise channel concatenation mechanism that operates on the channel dimension to better preserve appearance features. Our model delivers state-of-the-art video generation quality, generalizing from single-subject training to complex multi-subject scenarios with coherent synthesis and precise control over individual subjects, outperforming existing open-source and commercial baselines. To facilitate evaluation, we also introduce a comprehensive multi-subject video benchmark. Extensive experiments demonstrate the effectiveness of our approach, paving the way for scalable, controllable, and high-fidelity multi-subject video synthesis. Code and model can be found at: https://github.com/MAGREF-Video/MAGREF

Recent years have seen progress beyond domain-specific sound separation for speech or music towards universal sound separation for arbitrary sounds. Prior work on universal sound separation has investigated separating a target sound out of an audio mixture given a text query. Such text-queried sound separation systems provide a natural and scalable interface for specifying arbitrary target sounds. However, supervised text-queried sound separation systems require costly labeled audio-text pairs for training. Moreover, the audio provided in existing datasets is often recorded in a controlled environment, causing a considerable generalization gap to noisy audio in the wild. In this work, we aim to approach text-queried universal sound separation by using only unlabeled data. We propose to leverage the visual modality as a bridge to learn the desired audio-textual correspondence. The proposed CLIPSep model first encodes the input query into a query vector using the contrastive language-image pretraining (CLIP) model, and the query vector is then used to condition an audio separation model to separate out the target sound. While the model is trained on image-audio pairs extracted from unlabeled videos, at test time we can instead query the model with text inputs in a zero-shot setting, thanks to the joint language-image embedding learned by the CLIP model. Further, videos in the wild often contain off-screen sounds and background noise that may hinder the model from learning the desired audio-textual correspondence. To address this problem, we further propose an approach called noise invariant training for training a query-based sound separation model on noisy data. Experimental results show that the proposed models successfully learn text-queried universal sound separation using only noisy unlabeled videos, even achieving competitive performance against a supervised model in some settings.

Vision foundation models like DINOv2 demonstrate remarkable potential in medical imaging despite their origin in natural image domains. However, their design inherently works best for uni-modal image analysis, limiting their effectiveness for multi-modal imaging tasks that are common in many medical fields, such as neurology and oncology. While supervised models perform well in this setting, they fail to leverage unlabeled datasets and struggle with missing modalities, a frequent challenge in clinical settings. To bridge these gaps, we introduce MM-DINOv2, a novel and efficient framework that adapts the pre-trained vision foundation model DINOv2 for multi-modal medical imaging. Our approach incorporates multi-modal patch embeddings, enabling vision foundation models to effectively process multi-modal imaging data. To address missing modalities, we employ full-modality masking, which encourages the model to learn robust cross-modality relationships. Furthermore, we leverage semi-supervised learning to harness large unlabeled datasets, enhancing both the accuracy and reliability of medical predictions. Applied to glioma subtype classification from multi-sequence brain MRI, our method achieves a Matthews Correlation Coefficient (MCC) of 0.6 on an external test set, surpassing state-of-the-art supervised approaches by +11.1%. Our work establishes a scalable and robust solution for multi-modal medical imaging tasks, leveraging powerful vision foundation models pre-trained on natural images while addressing real-world clinical challenges such as missing data and limited annotations.

Although Contrastive Language-Image Pre-training (CLIP) exhibits strong performance across diverse vision tasks, its application to person representation learning faces two critical challenges: (i) the scarcity of large-scale annotated vision-language data focused on person-centric images, and (ii) the inherent limitations of global contrastive learning, which struggles to maintain discriminative local features crucial for fine-grained matching while remaining vulnerable to noisy text tokens. This work advances CLIP for person representation learning through synergistic improvements in data curation and model architecture. First, we develop a noise-resistant data construction pipeline that leverages the in-context learning capabilities of MLLMs to automatically filter and caption web-sourced images. This yields WebPerson, a large-scale dataset of 5M high-quality person-centric image-text pairs. Second, we introduce the GA-DMS (Gradient-Attention Guided Dual-Masking Synergetic) framework, which improves cross-modal alignment by adaptively masking noisy textual tokens based on the gradient-attention similarity score. Additionally, we incorporate masked token prediction objectives that compel the model to predict informative text tokens, enhancing fine-grained semantic representation learning. Extensive experiments show that GA-DMS achieves state-of-the-art performance across multiple benchmarks.

Instead of performing text-conditioned denoising in the image domain, latent diffusion models (LDMs) operate in latent space of a variational autoencoder (VAE), enabling more efficient processing at reduced computational costs. However, while the diffusion process has moved to the latent space, the contrastive language-image pre-training (CLIP) models, as used in many image processing tasks, still operate in pixel space. Doing so requires costly VAE-decoding of latent images before they can be processed. In this paper, we introduce Latent-CLIP, a CLIP model that operates directly in the latent space. We train Latent-CLIP on 2.7B pairs of latent images and descriptive texts, and show that it matches zero-shot classification performance of similarly sized CLIP models on both the ImageNet benchmark and a LDM-generated version of it, demonstrating its effectiveness in assessing both real and generated content. Furthermore, we construct Latent-CLIP rewards for reward-based noise optimization (ReNO) and show that they match the performance of their CLIP counterparts on GenEval and T2I-CompBench while cutting the cost of the total pipeline by 21%. Finally, we use Latent-CLIP to guide generation away from harmful content, achieving strong performance on the inappropriate image prompts (I2P) benchmark and a custom evaluation, without ever requiring the costly step of decoding intermediate images.

CLIP has achieved impressive zero-shot performance after pre-training on a large-scale dataset consisting of paired image-text data. Previous works have utilized CLIP by incorporating manually designed visual prompts like colored circles and blur masks into the images to guide the model's attention, showing enhanced zero-shot performance in downstream tasks. Although these methods have achieved promising results, they inevitably alter the original information of the images, which can lead to failure in specific tasks. We propose a train-free method Foveal-Attention CLIP (FALIP), which adjusts the CLIP's attention by inserting foveal attention masks into the multi-head self-attention module. We demonstrate FALIP effectively boosts CLIP zero-shot performance in tasks such as referring expressions comprehension, image classification, and 3D point cloud recognition. Experimental results further show that FALIP outperforms existing methods on most metrics and can augment current methods to enhance their performance.

Recent advances in generative models that iteratively synthesize audio clips sparked great success to text-to-audio synthesis (TTA), but with the cost of slow synthesis speed and heavy computation. Although there have been attempts to accelerate the iterative procedure, high-quality TTA systems remain inefficient due to hundreds of iterations required in the inference phase and large amount of model parameters. To address the challenges, we propose SpecMaskGIT, a light-weighted, efficient yet effective TTA model based on the masked generative modeling of spectrograms. First, SpecMaskGIT synthesizes a realistic 10s audio clip by less than 16 iterations, an order-of-magnitude less than previous iterative TTA methods.As a discrete model, SpecMaskGIT outperforms larger VQ-Diffusion and auto-regressive models in the TTA benchmark, while being real-time with only 4 CPU cores or even 30x faster with a GPU. Next, built upon a latent space of Mel-spectrogram, SpecMaskGIT has a wider range of applications (e.g., the zero-shot bandwidth extension) than similar methods built on the latent wave domain. Moreover, we interpret SpecMaskGIT as a generative extension to previous discriminative audio masked Transformers, and shed light on its audio representation learning potential. We hope our work inspires the exploration of masked audio modeling toward further diverse scenarios.

Multimodal information retrieval (MIR) faces inherent challenges due to the heterogeneity of data sources and the complexity of cross-modal alignment. While previous studies have identified modal gaps in feature spaces, a systematic approach to address these challenges remains unexplored. In this work, we introduce UNITE, a universal framework that tackles these challenges through two critical yet underexplored aspects: data curation and modality-aware training configurations. Our work provides the first comprehensive analysis of how modality-specific data properties influence downstream task performance across diverse scenarios. Moreover, we propose Modal-Aware Masked Contrastive Learning (MAMCL) to mitigate the competitive relationships among the instances of different modalities. Our framework achieves state-of-the-art results on multiple multimodal retrieval benchmarks, outperforming existing methods by notable margins. Through extensive experiments, we demonstrate that strategic modality curation and tailored training protocols are pivotal for robust cross-modal representation learning. This work not only advances MIR performance but also provides a foundational blueprint for future research in multimodal systems. Our project is available at https://friedrichor.github.io/projects/UNITE.

Image token removal is an efficient augmentation strategy for reducing the cost of computing image features. However, this efficient augmentation strategy has been found to adversely affect the accuracy of CLIP-based training. We hypothesize that removing a large portion of image tokens may improperly discard the semantic content associated with a given text description, thus constituting an incorrect pairing target in CLIP training. To address this issue, we propose an attentive token removal approach for CLIP training, which retains tokens with a high semantic correlation to the text description. The correlation scores are computed in an online fashion using the EMA version of the visual encoder. Our experiments show that the proposed attentive masking approach performs better than the previous method of random token removal for CLIP training. The approach also makes it efficient to apply multiple augmentation views to the image, as well as introducing instance contrastive learning tasks between these views into the CLIP framework. Compared to other CLIP improvements that combine different pre-training targets such as SLIP and MaskCLIP, our method is not only more effective, but also much more efficient. Specifically, using ViT-B and YFCC-15M dataset, our approach achieves 43.9% top-1 accuracy on ImageNet-1K zero-shot classification, as well as 62.7/42.1 and 38.0/23.2 I2T/T2I retrieval accuracy on Flickr30K and MS COCO, which are +1.1%, +5.5/+0.9, and +4.4/+1.3 higher than the SLIP method, while being 2.30times faster. An efficient version of our approach running 1.16times faster than the plain CLIP model achieves significant gains of +5.3%, +11.3/+8.0, and +9.5/+4.9 on these benchmarks.

In the past, the rapidly evolving field of sound classification greatly benefited from the application of methods from other domains. Today, we observe the trend to fuse domain-specific tasks and approaches together, which provides the community with new outstanding models. In this work, we present an extension of the CLIP model that handles audio in addition to text and images. Our proposed model incorporates the ESResNeXt audio-model into the CLIP framework using the AudioSet dataset. Such a combination enables the proposed model to perform bimodal and unimodal classification and querying, while keeping CLIP's ability to generalize to unseen datasets in a zero-shot inference fashion. AudioCLIP achieves new state-of-the-art results in the Environmental Sound Classification (ESC) task, out-performing other approaches by reaching accuracies of 90.07% on the UrbanSound8K and 97.15% on the ESC-50 datasets. Further it sets new baselines in the zero-shot ESC-task on the same datasets (68.78% and 69.40%, respectively). Finally, we also assess the cross-modal querying performance of the proposed model as well as the influence of full and partial training on the results. For the sake of reproducibility, our code is published.

Large sparsely-activated models have obtained excellent performance in multiple domains. However, such models are typically trained on a single modality at a time. We present the Language-Image MoE, LIMoE, a sparse mixture of experts model capable of multimodal learning. LIMoE accepts both images and text simultaneously, while being trained using a contrastive loss. MoEs are a natural fit for a multimodal backbone, since expert layers can learn an appropriate partitioning of modalities. However, new challenges arise; in particular, training stability and balanced expert utilization, for which we propose an entropy-based regularization scheme. Across multiple scales, we demonstrate remarkable performance improvement over dense models of equivalent computational cost. LIMoE-L/16 trained comparably to CLIP-L/14 achieves 78.6% zero-shot ImageNet accuracy (vs. 76.2%), and when further scaled to H/14 (with additional data) it achieves 84.1%, comparable to state-of-the-art methods which use larger custom per-modality backbones and pre-training schemes. We analyse the quantitative and qualitative behavior of LIMoE, and demonstrate phenomena such as differing treatment of the modalities and the organic emergence of modality-specific experts.