Scaling Concept With Text-Guided Diffusion Models

Text-guided diffusion models have revolutionized generative tasks by producing high-fidelity content from text descriptions. They have also enabled an editing paradigm where concepts can be replaced through text conditioning (e.g., a dog to a tiger). In this work, we explore a novel approach: instead of replacing a concept, can we enhance or suppress the concept itself? Through an empirical study, we identify a trend where concepts can be decomposed in text-guided diffusion models. Leveraging this insight, we introduce ScalingConcept, a simple yet effective method to scale decomposed concepts up or down in real input without introducing new elements. To systematically evaluate our approach, we present the WeakConcept-10 dataset, where concepts are imperfect and need to be enhanced. More importantly, ScalingConcept enables a variety of novel zero-shot applications across image and audio domains, including tasks such as canonical pose generation and generative sound highlighting or removal.

Will the Inclusion of Generated Data Amplify Bias Across Generations in Future Image Classification Models?

As the demand for high-quality training data escalates, researchers have increasingly turned to generative models to create synthetic data, addressing data scarcity and enabling continuous model improvement. However, reliance on self-generated data introduces a critical question: Will this practice amplify bias in future models? While most research has focused on overall performance, the impact on model bias, particularly subgroup bias, remains underexplored. In this work, we investigate the effects of the generated data on image classification tasks, with a specific focus on bias. We develop a practical simulation environment that integrates a self-consuming loop, where the generative model and classification model are trained synergistically. Hundreds of experiments are conducted on Colorized MNIST, CIFAR-20/100, and Hard ImageNet datasets to reveal changes in fairness metrics across generations. In addition, we provide a conjecture to explain the bias dynamics when training models on continuously augmented datasets across generations. Our findings contribute to the ongoing debate on the implications of synthetic data for fairness in real-world applications.

MMCOMPOSITION: Revisiting the Compositionality of Pre-trained Vision-Language Models

The advent of large Vision-Language Models (VLMs) has significantly advanced multimodal understanding, enabling more sophisticated and accurate integration of visual and textual information across various tasks, including image and video captioning, visual question answering, and cross-modal retrieval. Despite VLMs' superior capabilities, researchers lack a comprehensive understanding of their compositionality -- the ability to understand and produce novel combinations of known visual and textual components. Prior benchmarks provide only a relatively rough compositionality evaluation from the perspectives of objects, relations, and attributes while neglecting deeper reasoning about object interactions, counting, and complex compositions. However, compositionality is a critical ability that facilitates coherent reasoning and understanding across modalities for VLMs. To address this limitation, we propose MMCOMPOSITION, a novel human-annotated benchmark for comprehensively and accurately evaluating VLMs' compositionality. Our proposed benchmark serves as a complement to these earlier works. With MMCOMPOSITION, we can quantify and explore the compositionality of the mainstream VLMs. Surprisingly, we find GPT-4o's compositionality inferior to the best open-source model, and we analyze the underlying reasons. Our experimental analysis reveals the limitations of VLMs in fine-grained compositional perception and reasoning, and points to areas for improvement in VLM design and training. Resources available at: https://hanghuacs.github.io/MMComposition/

Language-Guided Joint Audio-Visual Editing via One-Shot Adaptation

In this paper, we introduce a novel task called language-guided joint audio-visual editing. Given an audio and image pair of a sounding event, this task aims at generating new audio-visual content by editing the given sounding event conditioned on the language guidance. For instance, we can alter the background environment of a sounding object while keeping its appearance unchanged, or we can add new sounds contextualized to the visual content. To address this task, we propose a new diffusion-based framework for joint audio-visual editing and introduce two key ideas. Firstly, we propose a one-shot adaptation approach to tailor generative diffusion models for audio-visual content editing. With as few as one audio-visual sample, we jointly transfer the audio and vision diffusion models to the target domain. After fine-tuning, our model enables consistent generation of this audio-visual sample. Secondly, we introduce a cross-modal semantic enhancement approach. We observe that when using language as content editing guidance, the vision branch may overlook editing requirements. This phenomenon, termed catastrophic neglect, hampers audio-visual alignment during content editing. We therefore enhance semantic consistency between language and vision to mitigate this issue. Extensive experiments validate the effectiveness of our method in language-based audio-visual editing and highlight its superiority over several baseline approaches. We recommend that readers visit our project page for more details: https://liangsusan-git.github.io/project/avedit/.

Quadratic Is Not What You Need For Multimodal Large Language Models

In the past year, the capabilities of Multimodal Large Language Models (MLLMs) have significantly improved across various aspects. However, constrained by the quadratic growth of computation in LLMs as the number of tokens increases, efficiency has become a bottleneck for further scaling MLLMs. Although recent efforts have been made to prune visual tokens or use more lightweight LLMs to reduce computation, the problem of quadratic growth in computation with the increase of visual tokens still persists. To address this, we propose a novel approach: instead of reducing the input visual tokens for LLMs, we focus on pruning vision-related computations within the LLMs. After pruning, the computation growth in the LLM is no longer quadratic with the increase of visual tokens, but linear. Surprisingly, we found that after applying such extensive pruning, the capabilities of MLLMs are comparable with the original one and even superior on some benchmarks with only 25% of the computation. This finding opens up the possibility for MLLMs to incorporate much denser visual tokens. Additionally, based on this finding, we further analyzed some architectural design deficiencies in existing MLLMs and proposed promising improvements. To the best of our knowledge, this is the first study to investigate the computational redundancy in the LLM's vision component of MLLMs. Code and checkpoints will be released soon.

EAGLE: Egocentric AGgregated Language-video Engine

The rapid evolution of egocentric video analysis brings new insights into understanding human activities and intentions from a first-person perspective. Despite this progress, the fragmentation in tasks like action recognition, procedure learning, and moment retrieval, \etc, coupled with inconsistent annotations and isolated model development, hinders a holistic interpretation of video content. In response, we introduce the EAGLE (Egocentric AGgregated Language-video Engine) model and the EAGLE-400K dataset to provide a unified framework that integrates various egocentric video understanding tasks. EAGLE-400K, the \textit{first} large-scale instruction-tuning dataset tailored for egocentric video, features 400K diverse samples to enhance a broad spectrum of tasks from activity recognition to procedure knowledge learning. Moreover, EAGLE, a strong video multimodal large language model (MLLM), is designed to effectively capture both spatial and temporal information. In addition, we propose a set of evaluation metrics designed to facilitate a thorough assessment of MLLM for egocentric video understanding. Our extensive experiments demonstrate EAGLE's superior performance over existing models, highlighting its ability to balance task-specific understanding with holistic video interpretation. With EAGLE, we aim to pave the way for research opportunities and practical applications in real-world scenarios.

CaRDiff: Video Salient Object Ranking Chain of Thought Reasoning for Saliency Prediction with Diffusion

Video saliency prediction aims to identify the regions in a video that attract human attention and gaze, driven by bottom-up features from the video and top-down processes like memory and cognition. Among these top-down influences, language plays a crucial role in guiding attention by shaping how visual information is interpreted. Existing methods primarily focus on modeling perceptual information while neglecting the reasoning process facilitated by language, where ranking cues are crucial outcomes of this process and practical guidance for saliency prediction. In this paper, we propose CaRDiff (Caption, Rank, and generate with Diffusion), a framework that imitates the process by integrating a multimodal large language model (MLLM), a grounding module, and a diffusion model, to enhance video saliency prediction. Specifically, we introduce a novel prompting method VSOR-CoT (Video Salient Object Ranking Chain of Thought), which utilizes an MLLM with a grounding module to caption video content and infer salient objects along with their rankings and positions. This process derives ranking maps that can be sufficiently leveraged by the diffusion model to decode the saliency maps for the given video accurately. Extensive experiments show the effectiveness of VSOR-CoT in improving the performance of video saliency prediction. The proposed CaRDiff performs better than state-of-the-art models on the MVS dataset and demonstrates cross-dataset capabilities on the DHF1k dataset through zero-shot evaluation.

Modeling and Driving Human Body Soundfields through Acoustic Primitives

While rendering and animation of photorealistic 3D human body models have matured and reached an impressive quality over the past years, modeling the spatial audio associated with such full body models has been largely ignored so far. In this work, we present a framework that allows for high-quality spatial audio generation, capable of rendering the full 3D soundfield generated by a human body, including speech, footsteps, hand-body interactions, and others. Given a basic audio-visual representation of the body in form of 3D body pose and audio from a head-mounted microphone, we demonstrate that we can render the full acoustic scene at any point in 3D space efficiently and accurately. To enable near-field and realtime rendering of sound, we borrow the idea of volumetric primitives from graphical neural rendering and transfer them into the acoustic domain. Our acoustic primitives result in an order of magnitude smaller soundfield representations and overcome deficiencies in near-field rendering compared to previous approaches.

Diversifying the Expert Knowledge for Task-Agnostic Pruning in Sparse Mixture-of-Experts

By increasing model parameters but activating them sparsely when performing a task, the use of Mixture-of-Experts (MoE) architecture significantly improves the performance of Large Language Models (LLMs) without increasing the inference cost. However, the memory consumption due to the growing number of experts presents a challenge to the deployment of these models in many real world settings. Our empirical study reveals that some experts encode redundant knowledge during pre-training. We thus propose a method of grouping and pruning similar experts to improve model's parameter efficiency. We validate the effectiveness of our method by pruning two state-of-the-art MoE models, Mixtral-8x7B and Mixtral-8x22B. Evaluation shows that our method outperforms other model pruning methods on a range of natural language tasks. To facilitate future research, we will release our code and the pruned MoE models.

Do More Details Always Introduce More Hallucinations in LVLM-based Image Captioning?

Large Vision-Language Models (LVLMs) excel in integrating visual and linguistic contexts to produce detailed content, facilitating applications such as image captioning. However, using LVLMs to generate descriptions often faces the challenge of object hallucination (OH), where the output text misrepresents actual objects in the input image. While previous studies attribute the occurrence of OH to the inclusion of more details, our study finds technical flaws in existing metrics, leading to unreliable evaluations of models and conclusions about OH. This has sparked a debate on the question: Do more details always introduce more hallucinations in LVLM-based image captioning? In this paper, we address this debate by proposing a novel decoding strategy, Differentiated Beam Decoding (DBD), along with a reliable new set of evaluation metrics: CLIP-Precision, CLIP-Recall, and CLIP-F1. DBD decodes the wealth of information hidden in visual input into distinct language representations called unit facts in parallel. This decoding is achieved via a well-designed differential score that guides the parallel search and candidate screening. The selected unit facts are then aggregated to generate the final caption. Our proposed metrics evaluate the comprehensiveness and accuracy of image captions by comparing the embedding groups of ground-truth image regions and generated text partitions. Extensive experiments on the Visual Genome dataset validate the effectiveness of our approach, demonstrating that it produces detailed descriptions while maintaining low hallucination levels.