We introduce Kosmos-2, a Multimodal Large Language Model (MLLM), enabling new capabilities of perceiving object descriptions (e.g., bounding boxes) and grounding text to the visual world. Specifically, we represent refer expressions as links in Markdown, i.e., ``[text span](bounding boxes)'', where object descriptions are sequences of location tokens. Together with multimodal corpora, we construct large-scale data of grounded image-text pairs (called GrIT) to train the model. In addition to the existing capabilities of MLLMs (e.g., perceiving general modalities, following instructions, and performing in-context learning), Kosmos-2 integrates the grounding capability into downstream applications. We evaluate Kosmos-2 on a wide range of tasks, including (i) multimodal grounding, such as referring expression comprehension, and phrase grounding, (ii) multimodal referring, such as referring expression generation, (iii) perception-language tasks, and (iv) language understanding and generation. This work lays out the foundation for the development of Embodiment AI and sheds light on the big convergence of language, multimodal perception, action, and world modeling, which is a key step toward artificial general intelligence. Code and pretrained models are available at //aka.ms/kosmos-2.
相關內容
The recent Long-Range Graph Benchmark (LRGB, Dwivedi et al. 2022) introduced a set of graph learning tasks strongly dependent on long-range interaction between vertices. Empirical evidence suggests that on these tasks Graph Transformers significantly outperform Message Passing GNNs (MPGNNs). In this paper, we carefully reevaluate multiple MPGNN baselines as well as the Graph Transformer GPS (Ramp\'a\v{s}ek et al. 2022) on LRGB. Through a rigorous empirical analysis, we demonstrate that the reported performance gap is overestimated due to suboptimal hyperparameter choices. It is noteworthy that across multiple datasets the performance gap completely vanishes after basic hyperparameter optimization. In addition, we discuss the impact of lacking feature normalization for LRGB's vision datasets and highlight a spurious implementation of LRGB's link prediction metric. The principal aim of our paper is to establish a higher standard of empirical rigor within the graph machine learning community.
Best Arm Identification (BAI) problems are progressively used for data-sensitive applications, such as designing adaptive clinical trials, tuning hyper-parameters, and conducting user studies to name a few. Motivated by the data privacy concerns invoked by these applications, we study the problem of BAI with fixed confidence under $\epsilon$-global Differential Privacy (DP). First, to quantify the cost of privacy, we derive a lower bound on the sample complexity of any $\delta$-correct BAI algorithm satisfying $\epsilon$-global DP. Our lower bound suggests the existence of two privacy regimes depending on the privacy budget $\epsilon$. In the high-privacy regime (small $\epsilon$), the hardness depends on a coupled effect of privacy and a novel information-theoretic quantity, called the Total Variation Characteristic Time. In the low-privacy regime (large $\epsilon$), the sample complexity lower bound reduces to the classical non-private lower bound. Second, we propose AdaP-TT, an $\epsilon$-global DP variant of the Top Two algorithm. AdaP-TT runs in arm-dependent adaptive episodes and adds Laplace noise to ensure a good privacy-utility trade-off. We derive an asymptotic upper bound on the sample complexity of AdaP-TT that matches with the lower bound up to multiplicative constants in the high-privacy regime. Finally, we provide an experimental analysis of AdaP-TT that validates our theoretical results.
In recent research, in the domain of speech processing, large End-to-End (E2E) systems for Automatic Speech Recognition (ASR) have reported state-of-the-art performance on various benchmarks. These systems intrinsically learn how to handle and remove noise conditions from speech. Previous research has shown, that it is possible to extract the denoising capabilities of these models into a preprocessor network, which can be used as a frontend for downstream ASR models. However, the proposed methods were limited to specific fully convolutional architectures. In this work, we propose a novel method to extract the denoising capabilities, that can be applied to any encoder-decoder architecture. We propose the Cleancoder preprocessor architecture that extracts hidden activations from the Conformer ASR model and feeds them to a decoder to predict denoised spectrograms. We train our pre-processor on the Noisy Speech Database (NSD) to reconstruct denoised spectrograms from noisy inputs. Then, we evaluate our model as a frontend to a pretrained Conformer ASR model as well as a frontend to train smaller Conformer ASR models from scratch. We show that the Cleancoder is able to filter noise from speech and that it improves the total Word Error Rate (WER) of the downstream model in noisy conditions for both applications.
Using accelerators based on analog computing is an efficient way to process the immensely large workloads in Neural Networks (NNs). One example of an analog computing scheme for NNs is Integrate-and-Fire (IF) Spiking Neural Networks (SNNs). However, to achieve high inference accuracy in IF-SNNs, the analog hardware needs to represent current-based multiply-accumulate (MAC) levels as spike times, for which a large membrane capacitor needs to be charged for a certain amount of time. A large capacitor results in high energy use, considerable area cost, and long latency, constituting one of the major bottlenecks in analog IF-SNN implementations. In this work, we propose a HW/SW Codesign method, called CapMin, for capacitor size minimization in analog computing IF-SNNs. CapMin minimizes the capacitor size by reducing the number of spike times needed for accurate operation of the HW, based on the absolute frequency of MAC level occurrences in the SW. To increase the operation of IF-SNNs to current variation, we propose the method CapMin-V, which trades capacitor size for protection based on the reduced capacitor size found in CapMin. In our experiments, CapMin achieves more than a 14$\times$ reduction in capacitor size over the state of the art, while CapMin-V achieves increased variation tolerance in the IF-SNN operation, requiring only a small increase in capacitor size.
This paper introduces the LDP-Auditor framework for empirically estimating the privacy loss of Locally Differentially Private (LDP) mechanisms. Several factors influencing the privacy audit are explored, such as the impact of different encoding and perturbation functions of eight state-of-the-art LDP protocols. Furthermore, the influence of domain size as well as the theoretical privacy loss parameter $\epsilon$ on local privacy estimation are also examined. Overall, our LDP-Auditor framework and findings offer valuable insights into the sources of randomness and information loss in LDP protocols, contributing to a more realistic understanding of the local privacy loss. Furthermore, we demonstrate the effectiveness of LDP-Auditor by successfully identifying a bug in an LDP library.
We study a higher-order surface finite-element (SFEM) penalty-based discretization of the tangential surface Stokes problem. Several discrete formulations are investigated which are equivalent in the continuous setting. The impact of the choice of discretization of the diffusion term and of the divergence term on numerical accuracy and convergence, as well as on implementation advantages, is discussed. We analyze the inf-sup stability of the discrete scheme in a generic approach by lifting stable finite-element pairs known from the literature. A discretization error analysis in tangential norms then shows optimal order convergence of an isogeometric setting that requires only geometric knowledge of the discrete surface.
Deep Learning (DL) frameworks are now widely used, simplifying the creation of complex models as well as their integration to various applications even to non DL experts. However, like any other programs, they are prone to bugs. This paper deals with the subcategory of bugs named silent bugs: they lead to wrong behavior but they do not cause system crashes or hangs, nor show an error message to the user. Such bugs are even more dangerous in DL applications and frameworks due to the "black-box" and stochastic nature of the systems (the end user can not understand how the model makes decisions). This paper presents the first empirical study of Keras and TensorFlow silent bugs, and their impact on users' programs. We extracted closed issues related to Keras from the TensorFlow GitHub repository. Out of the 1,168 issues that we gathered, 77 were reproducible silent bugs affecting users' programs. We categorized the bugs based on the effects on the users' programs and the components where the issues occurred, using information from the issue reports. We then derived a threat level for each of the issues, based on the impact they had on the users' programs. To assess the relevance of identified categories and the impact scale, we conducted an online survey with 103 DL developers. The participants generally agreed with the significant impact of silent bugs in DL libraries and acknowledged our findings (i.e., categories of silent bugs and the proposed impact scale). Finally, leveraging our analysis, we provide a set of guidelines to facilitate safeguarding against such bugs in DL frameworks.
Automation of High-Level Context (HLC) reasoning for intelligent systems at scale is imperative due to the unceasing accumulation of contextual data in the IoT era, the trend of the fusion of data from multi-sources, and the intrinsic complexity and dynamism of the context-based decision-making process. To mitigate this issue, we propose an automatic context reasoning framework CSM-H-R, which programmatically combines ontologies and states at runtime and the model-storage phase for attaining the ability to recognize meaningful HLC, and the resulting data representation can be applied to different reasoning techniques. Case studies are developed based on an intelligent elevator system in a smart campus setting. An implementation of the framework - a CSM Engine, and the experiments of translating the HLC reasoning into vector and matrix computing especially take care of the dynamic aspects of context and present the potentiality of using advanced mathematical and probabilistic models to achieve the next level of automation in integrating intelligent systems; meanwhile, privacy protection support is achieved by anonymization through label embedding and reducing information correlation. The code of this study is available at: //github.com/songhui01/CSM-H-R.
Object proposal generation serves as a standard pre-processing step in Vision-Language (VL) tasks (image captioning, visual question answering, etc.). The performance of object proposals generated for VL tasks is currently evaluated across all available annotations, a protocol that we show is misaligned - higher scores do not necessarily correspond to improved performance on downstream VL tasks. Our work serves as a study of this phenomenon and explores the effectiveness of semantic grounding to mitigate its effects. To this end, we propose evaluating object proposals against only a subset of available annotations, selected by thresholding an annotation importance score. Importance of object annotations to VL tasks is quantified by extracting relevant semantic information from text describing the image. We show that our method is consistent and demonstrates greatly improved alignment with annotations selected by image captioning metrics and human annotation when compared against existing techniques. Lastly, we compare current detectors used in the Scene Graph Generation (SGG) benchmark as a use case, which serves as an example of when traditional object proposal evaluation techniques are misaligned.
The problem of Multiple Object Tracking (MOT) consists in following the trajectory of different objects in a sequence, usually a video. In recent years, with the rise of Deep Learning, the algorithms that provide a solution to this problem have benefited from the representational power of deep models. This paper provides a comprehensive survey on works that employ Deep Learning models to solve the task of MOT on single-camera videos. Four main steps in MOT algorithms are identified, and an in-depth review of how Deep Learning was employed in each one of these stages is presented. A complete experimental comparison of the presented works on the three MOTChallenge datasets is also provided, identifying a number of similarities among the top-performing methods and presenting some possible future research directions.
小貼士
登錄享
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191