Score distillation sampling (SDS) and its variants have greatly boosted the development of text-to-3D generation, but are vulnerable to geometry collapse and poor textures yet. To solve this issue, we first deeply analyze the SDS and find that its distillation sampling process indeed corresponds to the trajectory sampling of a stochastic differential equation (SDE): SDS samples along an SDE trajectory to yield a less noisy sample which then serves as a guidance to optimize a 3D model. However, the randomness in SDE sampling often leads to a diverse and unpredictable sample which is not always less noisy, and thus is not a consistently correct guidance, explaining the vulnerability of SDS. Since for any SDE, there always exists an ordinary differential equation (ODE) whose trajectory sampling can deterministically and consistently converge to the desired target point as the SDE, we propose a novel and effective "Consistent3D" method that explores the ODE deterministic sampling prior for text-to-3D generation. Specifically, at each training iteration, given a rendered image by a 3D model, we first estimate its desired 3D score function by a pre-trained 2D diffusion model, and build an ODE for trajectory sampling. Next, we design a consistency distillation sampling loss which samples along the ODE trajectory to generate two adjacent samples and uses the less noisy sample to guide another more noisy one for distilling the deterministic prior into the 3D model. Experimental results show the efficacy of our Consistent3D in generating high-fidelity and diverse 3D objects and large-scale scenes, as shown in Fig. 1. The codes are available at https://github.com/sail-sg/Consistent3D.
We introduce the concept of multiple temporal perspectives, a novel approach applicable to Recurrent Neural Network (RNN) architectures for enhancing their understanding of sequential data. This method involves maintaining diverse temporal views of previously encountered text, significantly enriching the language models' capacity to interpret context. To show the efficacy of this approach, we incorporate it into the Receptance Weighted Key Value (RWKV) architecture, addressing its inherent challenge of retaining all historical information within a single hidden state. Notably, this improvement is achieved with a minimal increase in the number of parameters --even as little as $0.04\%$ of the original number of parameters. Further, the additional parameters necessary for the multiple temporal perspectives are fine-tuned with minimal computational overhead, avoiding the need for a full pre-training. The resulting model maintains linear computational complexity during prompt inference, ensuring consistent efficiency across various sequence lengths. The empirical results and ablation studies included in our research validate the effectiveness of our approach, showcasing improved performance across multiple benchmarks. The code, model weights and datasets are open-sourced at: https://github.com/RazvanDu/TemporalRNNs.
In this work we study the enhancement of Low Rank Adaptation (LoRA) fine-tuning procedure by introducing a Riemannian preconditioner in its optimization step. Specifically, we introduce an $r\times r$ preconditioner in each gradient step where $r$ is the LoRA rank. This preconditioner requires a small change to existing optimizer code and creates virtually minuscule storage and runtime overhead. Our experimental results with both large language models and text-to-image diffusion models show that with our preconditioner, the convergence and reliability of SGD and AdamW can be significantly enhanced. Moreover, the training process becomes much more robust to hyperparameter choices such as learning rate. Theoretically, we show that fine-tuning a two-layer ReLU network in the convex paramaterization with our preconditioner has convergence rate independent of condition number of the data matrix. This new Riemannian preconditioner, previously explored in classic low-rank matrix recovery, is introduced to deep learning tasks for the first time in our work. We release our code at https://github.com/pilancilab/Riemannian_Preconditioned_LoRA.
Generative AI has witnessed rapid advancement in recent years, expanding their capabilities to create synthesized content such as text, images, audio, and code. The high fidelity and authenticity of contents generated by these Deep Generative Models (DGMs) have sparked significant copyright concerns. There have been various legal debates on how to effectively safeguard copyrights in DGMs. This work delves into this issue by providing a comprehensive overview of copyright protection from a technical perspective. We examine from two distinct viewpoints: the copyrights pertaining to the source data held by the data owners and those of the generative models maintained by the model builders. For data copyright, we delve into methods data owners can protect their content and DGMs can be utilized without infringing upon these rights. For model copyright, our discussion extends to strategies for preventing model theft and identifying outputs generated by specific models. Finally, we highlight the limitations of existing techniques and identify areas that remain unexplored. Furthermore, we discuss prospective directions for the future of copyright protection, underscoring its importance for the sustainable and ethical development of Generative AI.
This paper explores a specific sub-task of cross-modal music retrieval. We consider the delicate task of retrieving a performance or rendition of a musical piece based on a description of its style, expressive character, or emotion from a set of different performances of the same piece. We observe that a general purpose cross-modal system trained to learn a common text-audio embedding space does not yield optimal results for this task. By introducing two changes -- one each to the text encoder and the audio encoder -- we demonstrate improved performance on a dataset of piano performances and associated free-text descriptions. On the text side, we use emotion-enriched word embeddings (EWE) and on the audio side, we extract mid-level perceptual features instead of generic audio embeddings. Our results highlight the effectiveness of mid-level perceptual features learnt from music and emotion enriched word embeddings learnt from emotion-labelled text in capturing musical expression in a cross-modal setting. Additionally, our interpretable mid-level features provide a route for introducing explainability in the retrieval and downstream recommendation processes.
Social media accounts post increasingly similar content, creating a chaotic experience across platforms, which makes accessing desired information difficult. These posts can be organized by categorizing and grouping duplicates across social handles and accounts. There can be more than one duplicate of a post, however, a conventional Siamese neural network only considers a pair of inputs for duplicate text detection. In this paper, we first propose a multiple-input Siamese network, MultiSiam. This condensed network is then used to propose another model, SMCD (Social Media Classification and Duplication Model) to perform both duplicate text grouping and categorization. The MultiSiam network, just like the Siamese, can be used in multiple applications by changing the sub-network appropriately.
Within the scope of natural language processing, the domain of multi-label text classification is uniquely challenging due to its expansive and uneven label distribution. The complexity deepens due to the demand for an extensive set of annotated data for training an advanced deep learning model, especially in specialized fields where the labeling task can be labor-intensive and often requires domain-specific knowledge. Addressing these challenges, our study introduces a novel deep active learning strategy, capitalizing on the Beta family of proper scoring rules within the Expected Loss Reduction framework. It computes the expected increase in scores using the Beta Scoring Rules, which are then transformed into sample vector representations. These vector representations guide the diverse selection of informative samples, directly linking this process to the model's expected proper score. Comprehensive evaluations across both synthetic and real datasets reveal our method's capability to often outperform established acquisition techniques in multi-label text classification, presenting encouraging outcomes across various architectural and dataset scenarios.
Vision-Language Models (VLMs) have emerged as the state-of-the-art representation learning solution, with myriads of downstream applications such as image classification, retrieval and generation. A natural question is whether these models memorize their training data, which also has implications for generalization. We propose a new method for measuring memorization in VLMs, which we call d\'ej\`a vu memorization. For VLMs trained on image-caption pairs, we show that the model indeed retains information about individual objects in the training images beyond what can be inferred from correlations or the image caption. We evaluate d\'ej\`a vu memorization at both sample and population level, and show that it is significant for OpenCLIP trained on as many as 50M image-caption pairs. Finally, we show that text randomization considerably mitigates memorization while only moderately impacting the model's downstream task performance.
Over the past decade, visual gaze estimation has garnered growing attention within the research community, thanks to its wide-ranging application scenarios. While existing estimation approaches have achieved remarkable success in enhancing prediction accuracy, they primarily infer gaze directions from single-image signals and discard the huge potentials of the currently dominant text guidance. Notably, visual-language collaboration has been extensively explored across a range of visual tasks, such as image synthesis and manipulation, leveraging the remarkable transferability of large-scale Contrastive Language-Image Pre-training (CLIP) model. Nevertheless, existing gaze estimation approaches ignore the rich semantic cues conveyed by linguistic signals and priors in CLIP feature space, thereby yielding performance setbacks. In pursuit of making up this gap, we delve deeply into the text-eye collaboration protocol and introduce a novel gaze estimation framework in this paper, referred to as GazeCLIP. Specifically, we intricately design a linguistic description generator to produce text signals with coarse directional cues. Additionally, a CLIP-based backbone that excels in characterizing text-eye pairs for gaze estimation is presented. This is followed by the implementation of a fine-grained multi-modal fusion module aimed at modeling the interrelationships between heterogeneous inputs. Extensive experiments on three challenging datasets demonstrate the superiority of the proposed GazeCLIP which surpasses the previous approaches and achieves the state-of-the-art estimation accuracy.
Most advances in medical image recognition supporting clinical auxiliary diagnosis meet challenges due to the low-resource situation in the medical field, where annotations are highly expensive and professional. This low-resource problem can be alleviated by leveraging the transferable representations of large-scale pre-trained vision-language models via relevant medical text prompts. However, existing pre-trained vision-language models require domain experts to carefully design the medical prompts, which greatly increases the burden on clinicians. To address this problem, we propose a weakly supervised prompt learning method MedPrompt to automatically generate medical prompts, which includes an unsupervised pre-trained vision-language model and a weakly supervised prompt learning model. The unsupervised pre-trained vision-language model utilizes the natural correlation between medical images and corresponding medical texts for pre-training, without any manual annotations. The weakly supervised prompt learning model only utilizes the classes of images in the dataset to guide the learning of the specific class vector in the prompt, while the learning of other context vectors in the prompt requires no manual annotations for guidance. To the best of our knowledge, this is the first model to automatically generate medical prompts. With these prompts, the pre-trained vision-language model can be freed from the strong expert dependency of manual annotation and manual prompt design. Experimental results show that the model using our automatically generated prompts outperforms its full-shot learning hand-crafted prompts counterparts with only a minimal number of labeled samples for few-shot learning, and reaches superior or comparable accuracy on zero-shot image classification. The proposed prompt generator is lightweight and therefore can be embedded into any network architecture.