Large language models have unlocked strong multi-task capabilities from reading instructive prompts. However, recent studies have shown that existing large models still have difficulty with information extraction tasks. For example, gpt-3.5-turbo achieved an F1 score of 18.22 on the Ontonotes dataset, which is significantly lower than the state-of-the-art performance. In this paper, we propose InstructUIE, a unified information extraction framework based on instruction tuning, which can uniformly model various information extraction tasks and capture the inter-task dependency. To validate the proposed method, we introduce IE INSTRUCTIONS, a benchmark of 32 diverse information extraction datasets in a unified text-to-text format with expert-written instructions. Experimental results demonstrate that our method achieves comparable performance to Bert in supervised settings and significantly outperforms the state-of-the-art and gpt3.5 in zero-shot settings.
This paper explores the potential of curriculum learning in LiDAR-based 3D object detection by proposing a curricular object manipulation (COM) framework. The framework embeds the curricular training strategy into both the loss design and the augmentation process. For the loss design, we propose the COMLoss to dynamically predict object-level difficulties and emphasize objects of different difficulties based on training stages. On top of the widely-used augmentation technique called GT-Aug in LiDAR detection tasks, we propose a novel COMAug strategy which first clusters objects in ground-truth database based on well-designed heuristics. Group-level difficulties rather than individual ones are then predicted and updated during training for stable results. Model performance and generalization capabilities can be improved by sampling and augmenting progressively more difficult objects into the training samples. Extensive experiments and ablation studies reveal the superior and generality of the proposed framework. The code is available at https://github.com/ZZY816/COM.
Sampling from high-dimensional distributions is a fundamental problem in statistical research and practice. However, great challenges emerge when the target density function is unnormalized and contains isolated modes. We tackle this difficulty by fitting an invertible transformation mapping, called a transport map, between a reference probability measure and the target distribution, so that sampling from the target distribution can be achieved by pushing forward a reference sample through the transport map. We theoretically analyze the limitations of existing transport-based sampling methods using the Wasserstein gradient flow theory, and propose a new method called TemperFlow that addresses the multimodality issue. TemperFlow adaptively learns a sequence of tempered distributions to progressively approach the target distribution, and we prove that it overcomes the limitations of existing methods. Various experiments demonstrate the superior performance of this novel sampler compared to traditional methods, and we show its applications in modern deep learning tasks such as image generation. The programming code for the numerical experiments is available at https://github.com/yixuan/temperflow.
There has been a recent explosion of computer vision models which perform many tasks and are composed of an image encoder (usually a ViT) and an autoregressive decoder (usually a Transformer). However, most of this work simply presents one system and its results, leaving many questions regarding design decisions and trade-offs of such systems unanswered. In this work, we aim to provide such answers. We take a close look at autoregressive decoders for multi-task learning in multimodal computer vision, including classification, captioning, visual question answering, and optical character recognition. Through extensive systematic experiments, we study the effects of task and data mixture, training and regularization hyperparameters, conditioning type and specificity, modality combination, and more. Importantly, we compare these to well-tuned single-task baselines to highlight the cost incurred by multi-tasking. A key finding is that a small decoder learned on top of a frozen pretrained encoder works surprisingly well. We call this setup locked-image tuning with decoder (LiT-decoder). It can be seen as teaching a decoder to interact with a pretrained vision model via natural language.
Existing Transformer-based RGBT tracking methods either use cross-attention to fuse the two modalities, or use self-attention and cross-attention to model both modality-specific and modality-sharing information. However, the significant appearance gap between modalities limits the feature representation ability of certain modalities during the fusion process. To address this problem, we propose a novel Progressive Fusion Transformer called ProFormer, which progressively integrates single-modality information into the multimodal representation for robust RGBT tracking. In particular, ProFormer first uses a self-attention module to collaboratively extract the multimodal representation, and then uses two cross-attention modules to interact it with the features of the dual modalities respectively. In this way, the modality-specific information can well be activated in the multimodal representation. Finally, a feed-forward network is used to fuse two interacted multimodal representations for the further enhancement of the final multimodal representation. In addition, existing learning methods of RGBT trackers either fuse multimodal features into one for final classification, or exploit the relationship between unimodal branches and fused branch through a competitive learning strategy. However, they either ignore the learning of single-modality branches or result in one branch failing to be well optimized. To solve these problems, we propose a dynamically guided learning algorithm that adaptively uses well-performing branches to guide the learning of other branches, for enhancing the representation ability of each branch. Extensive experiments demonstrate that our proposed ProFormer sets a new state-of-the-art performance on RGBT210, RGBT234, LasHeR, and VTUAV datasets.
The last decade has witnessed the proliferation of micro-videos on various user-generated content platforms. According to our statistics, around 85.7\% of micro-videos lack annotation. In this paper, we focus on annotating micro-videos with tags. Existing methods mostly focus on analyzing video content, neglecting users' social influence and tag relation. Meanwhile, existing tag relation construction methods suffer from either deficient performance or low tag coverage. To jointly model social influence and tag relation, we formulate micro-video tagging as a link prediction problem in a constructed heterogeneous network. Specifically, the tag relation (represented by tag ontology) is constructed in a semi-supervised manner. Then, we combine tag relation, video-tag annotation, and user-follow relation to build the network. Afterward, a better video and tag representation are derived through Behavior Spread modeling and visual and linguistic knowledge aggregation. Finally, the semantic similarity between each micro-video and all candidate tags is calculated in this video-tag network. Extensive experiments on industrial datasets of three verticals verify the superiority of our model compared with several state-of-the-art baselines.
Unsupervised learning of vision transformers seeks to pretrain an encoder via pretext tasks without labels. Among them is the Masked Image Modeling (MIM) aligned with pretraining of language transformers by predicting masked patches as a pretext task. A criterion in unsupervised pretraining is the pretext task needs to be sufficiently hard to prevent the transformer encoder from learning trivial low-level features not generalizable well to downstream tasks. For this purpose, we propose an Adversarial Positional Embedding (AdPE) approach -- It distorts the local visual structures by perturbing the position encodings so that the learned transformer cannot simply use the locally correlated patches to predict the missing ones. We hypothesize that it forces the transformer encoder to learn more discriminative features in a global context with stronger generalizability to downstream tasks. We will consider both absolute and relative positional encodings, where adversarial positions can be imposed both in the embedding mode and the coordinate mode. We will also present a new MAE+ baseline that brings the performance of the MIM pretraining to a new level with the AdPE. The experiments demonstrate that our approach can improve the fine-tuning accuracy of MAE by $0.8\%$ and $0.4\%$ over 1600 epochs of pretraining ViT-B and ViT-L on Imagenet1K. For the transfer learning task, it outperforms the MAE with the ViT-B backbone by $2.6\%$ in mIoU on ADE20K, and by $3.2\%$ in AP$^{bbox}$ and $1.6\%$ in AP$^{mask}$ on COCO, respectively. These results are obtained with the AdPE being a pure MIM approach that does not use any extra models or external datasets for pretraining. The code is available at https://github.com/maple-research-lab/AdPE.
With the urgent demand for generalized deep models, many pre-trained big models are proposed, such as BERT, ViT, GPT, etc. Inspired by the success of these models in single domains (like computer vision and natural language processing), the multi-modal pre-trained big models have also drawn more and more attention in recent years. In this work, we give a comprehensive survey of these models and hope this paper could provide new insights and helps fresh researchers to track the most cutting-edge works. Specifically, we firstly introduce the background of multi-modal pre-training by reviewing the conventional deep learning, pre-training works in natural language process, computer vision, and speech. Then, we introduce the task definition, key challenges, and advantages of multi-modal pre-training models (MM-PTMs), and discuss the MM-PTMs with a focus on data, objectives, network architectures, and knowledge enhanced pre-training. After that, we introduce the downstream tasks used for the validation of large-scale MM-PTMs, including generative, classification, and regression tasks. We also give visualization and analysis of the model parameters and results on representative downstream tasks. Finally, we point out possible research directions for this topic that may benefit future works. In addition, we maintain a continuously updated paper list for large-scale pre-trained multi-modal big models: https://github.com/wangxiao5791509/MultiModal_BigModels_Survey
Graph neural networks (GNNs) have attracted considerable attention from the research community. It is well established that GNNs are usually roughly divided into spatial and spectral methods. Despite that spectral GNNs play an important role in both graph signal processing and graph representation learning, existing studies are biased toward spatial approaches, and there is no comprehensive review on spectral GNNs so far. In this paper, we summarize the recent development of spectral GNNs, including model, theory, and application. Specifically, we first discuss the connection between spatial GNNs and spectral GNNs, which shows that spectral GNNs can capture global information and have better expressiveness and interpretability. Next, we categorize existing spectral GNNs according to the spectrum information they use, \ie, eigenvalues or eigenvectors. In addition, we review major theoretical results and applications of spectral GNNs, followed by a quantitative experiment to benchmark some popular spectral GNNs. Finally, we conclude the paper with some future directions.
The scaling of Transformers has driven breakthrough capabilities for language models. At present, the largest large language models (LLMs) contain upwards of 100B parameters. Vision Transformers (ViT) have introduced the same architecture to image and video modelling, but these have not yet been successfully scaled to nearly the same degree; the largest dense ViT contains 4B parameters (Chen et al., 2022). We present a recipe for highly efficient and stable training of a 22B-parameter ViT (ViT-22B) and perform a wide variety of experiments on the resulting model. When evaluated on downstream tasks (often with a lightweight linear model on frozen features), ViT-22B demonstrates increasing performance with scale. We further observe other interesting benefits of scale, including an improved tradeoff between fairness and performance, state-of-the-art alignment to human visual perception in terms of shape/texture bias, and improved robustness. ViT-22B demonstrates the potential for "LLM-like" scaling in vision, and provides key steps towards getting there.