Localizing objects in 3D scenes according to the semantics of a given natural language is a fundamental yet important task in the field of multimedia understanding, which benefits various real-world applications such as robotics and autonomous driving. However, the majority of existing 3D object grounding methods are restricted to a single-sentence input describing an individual object, which cannot comprehend and reason more contextualized descriptions of multiple objects in more practical 3D cases. To this end, we introduce a new challenging task, called 3D Dense Object Grounding (3D DOG), to jointly localize multiple objects described in a more complicated paragraph rather than a single sentence. Instead of naively localizing each sentence-guided object independently, we found that dense objects described in the same paragraph are often semantically related and spatially located in a focused region of the 3D scene. To explore such semantic and spatial relationships of densely referred objects for more accurate localization, we propose a novel Stacked Transformer based framework for 3D DOG, named 3DOGSFormer. Specifically, we first devise a contextual query-driven local transformer decoder to generate initial grounding proposals for each target object. Then, we employ a proposal-guided global transformer decoder that exploits the local object features to learn their correlation for further refining initial grounding proposals. Extensive experiments on three challenging benchmarks (Nr3D, Sr3D, and ScanRefer) show that our proposed 3DOGSFormer outperforms state-of-the-art 3D single-object grounding methods and their dense-object variants by significant margins.
相關內容
Reasoning over spans of tokens from different parts of the input is essential for natural language understanding (NLU) tasks such as fact-checking (FC), machine reading comprehension (MRC) or natural language inference (NLI). However, existing highlight-based explanations primarily focus on identifying individual important tokens or interactions only between adjacent tokens or tuples of tokens. Most notably, there is a lack of annotations capturing the human decision-making process w.r.t. the necessary interactions for informed decision-making in such tasks. To bridge this gap, we introduce SpanEx, a multi-annotator dataset of human span interaction explanations for two NLU tasks: NLI and FC. We then investigate the decision-making processes of multiple fine-tuned large language models in terms of the employed connections between spans in separate parts of the input and compare them to the human reasoning processes. Finally, we present a novel community detection based unsupervised method to extract such interaction explanations from a model's inner workings.
Multimodal large language models (MLLMs) have shown great potential in perception and interpretation tasks, but their capabilities in predictive reasoning remain under-explored. To address this gap, we introduce a novel benchmark that assesses the predictive reasoning capabilities of MLLMs across diverse scenarios. Our benchmark targets three important domains: abstract pattern reasoning, human activity prediction, and physical interaction prediction. We further develop three evaluation methods powered by large language model to robustly quantify a model's performance in predicting and reasoning the future based on multi-visual context. Empirical experiments confirm the soundness of the proposed benchmark and evaluation methods via rigorous testing and reveal pros and cons of current popular MLLMs in the task of predictive reasoning. Lastly, our proposed benchmark provides a standardized evaluation framework for MLLMs and can facilitate the development of more advanced models that can reason and predict over complex long sequence of multimodal input.
Modern neural language models (LMs) are powerful tools for modeling human sentence production and comprehension, and their internal representations are remarkably well-aligned with representations of language in the human brain. But to achieve these results, LMs must be trained in distinctly un-human-like ways -- requiring orders of magnitude more language data than children receive during development, and without any of the accompanying grounding in perception, action, or social behavior. Do models trained more naturalistically -- with grounded supervision -- exhibit more human-like language learning? We investigate this question in the context of word learning, a key sub-task in language acquisition. We train a diverse set of LM architectures, with and without auxiliary supervision from image captioning tasks, on datasets of varying scales. We then evaluate these models on a broad set of benchmarks characterizing models' learning of syntactic categories, lexical relations, semantic features, semantic similarity, and alignment with human neural representations. We find that visual supervision can indeed improve the efficiency of word learning. However, these improvements are limited: they are present almost exclusively in the low-data regime, and sometimes canceled out by the inclusion of rich distributional signals from text. The information conveyed by text and images is not redundant -- we find that models mainly driven by visual information yield qualitatively different from those mainly driven by word co-occurrences. However, our results suggest that current multi-modal modeling approaches fail to effectively leverage visual information to build more human-like word representations from human-sized datasets.
Radar odometry estimation has emerged as a critical technique in the field of autonomous navigation, providing robust and reliable motion estimation under various environmental conditions. Despite its potential, the complex nature of radar signals and the inherent challenges associated with processing these signals have limited the widespread adoption of this technology. This paper aims to address these challenges by proposing novel improvements to an existing method for radar odometry estimation, designed to enhance accuracy and reliability in diverse scenarios. Our pipeline consists of filtering, motion compensation, oriented surface points computation, smoothing, one-to-many radar scan registration, and pose refinement. The developed method enforces local understanding of the scene, by adding additional information through smoothing techniques, and alignment of consecutive scans, as a refinement posterior to the one-to-many registration. We present an in-depth investigation of the contribution of each improvement to the localization accuracy, and we benchmark our system on the sequences of the main datasets for radar understanding, i.e., the Oxford Radar RobotCar, MulRan, and Boreas datasets. The proposed pipeline is able to achieve superior results, on all scenarios considered and under harsh environmental constraints.
Bayesian hypothesis testing leverages posterior probabilities, Bayes factors, or credible intervals to assess characteristics that summarize data. We propose a framework for power curve approximation with such hypothesis tests that assumes data are generated using statistical models with fixed parameters for the purposes of sample size determination. We present a fast approach to explore the sampling distribution of posterior probabilities when the conditions for the Bernstein-von Mises theorem are satisfied. We extend that approach to facilitate targeted sampling from the approximate sampling distribution of posterior probabilities for each sample size explored. These sampling distributions are used to construct power curves for various types of posterior analyses. Our resulting method for power curve approximation is orders of magnitude faster than conventional power curve estimation for Bayesian hypothesis tests. We also prove the consistency of the corresponding power estimates and sample size recommendations under certain conditions.
Significant computational resources are required to train Graph Neural Networks (GNNs) at a large scale, and the process is highly data-intensive. One of the most effective ways to reduce resource requirements is minibatch training coupled with graph sampling. GNNs have the unique property that items in a minibatch have overlapping data. However, the commonly implemented Independent Minibatching approach assigns each Processing Element (PE) its own minibatch to process, leading to duplicated computations and input data access across PEs. This amplifies the Neighborhood Explosion Phenomenon (NEP), which is the main bottleneck limiting scaling. To reduce the effects of NEP in the multi-PE setting, we propose a new approach called Cooperative Minibatching. Our approach capitalizes on the fact that the size of the sampled subgraph is a concave function of the batch size, leading to significant reductions in the amount of work per seed vertex as batch sizes increase. Hence, it is favorable for processors equipped with a fast interconnect to work on a large minibatch together as a single larger processor, instead of working on separate smaller minibatches, even though global batch size is identical. We also show how to take advantage of the same phenomenon in serial execution by generating dependent consecutive minibatches. Our experimental evaluations show up to 4x bandwidth savings for fetching vertex embeddings, by simply increasing this dependency without harming model convergence. Combining our proposed approaches, we achieve up to 64% speedup over Independent Minibatching on single-node multi-GPU systems.
Security properties are often focused on the technological side of the system. One implicitly assumes that the users will behave in the right way to preserve the property at hand. In real life, this cannot be taken for granted. In particular, security mechanisms that are difficult and costly to use are often ignored by the users, and do not really defend the system against possible attacks. Here, we propose a graded notion of security based on the complexity of the user's strategic behavior. More precisely, we suggest that the level to which a security property $\varphi$ is satisfied can be defined in terms of (a) the complexity of the strategy that the voter needs to execute to make $\varphi$ true, and (b) the resources that the user must employ on the way. The simpler and cheaper to obtain $\varphi$, the higher the degree of security. We demonstrate how the idea works in a case study based on an electronic voting scenario. To this end, we model the vVote implementation of the \Pret voting protocol for coercion-resistant and voter-verifiable elections. Then, we identify "natural" strategies for the voter to obtain receipt-freeness, and measure the voter's effort that they require. We also look at how hard it is for the coercer to compromise the election through a randomization attack.
Text to speech (TTS), or speech synthesis, which aims to synthesize intelligible and natural speech given text, is a hot research topic in speech, language, and machine learning communities and has broad applications in the industry. As the development of deep learning and artificial intelligence, neural network-based TTS has significantly improved the quality of synthesized speech in recent years. In this paper, we conduct a comprehensive survey on neural TTS, aiming to provide a good understanding of current research and future trends. We focus on the key components in neural TTS, including text analysis, acoustic models and vocoders, and several advanced topics, including fast TTS, low-resource TTS, robust TTS, expressive TTS, and adaptive TTS, etc. We further summarize resources related to TTS (e.g., datasets, opensource implementations) and discuss future research directions. This survey can serve both academic researchers and industry practitioners working on TTS.
The inductive biases of graph representation learning algorithms are often encoded in the background geometry of their embedding space. In this paper, we show that general directed graphs can be effectively represented by an embedding model that combines three components: a pseudo-Riemannian metric structure, a non-trivial global topology, and a unique likelihood function that explicitly incorporates a preferred direction in embedding space. We demonstrate the representational capabilities of this method by applying it to the task of link prediction on a series of synthetic and real directed graphs from natural language applications and biology. In particular, we show that low-dimensional cylindrical Minkowski and anti-de Sitter spacetimes can produce equal or better graph representations than curved Riemannian manifolds of higher dimensions.
Embedding entities and relations into a continuous multi-dimensional vector space have become the dominant method for knowledge graph embedding in representation learning. However, most existing models ignore to represent hierarchical knowledge, such as the similarities and dissimilarities of entities in one domain. We proposed to learn a Domain Representations over existing knowledge graph embedding models, such that entities that have similar attributes are organized into the same domain. Such hierarchical knowledge of domains can give further evidence in link prediction. Experimental results show that domain embeddings give a significant improvement over the most recent state-of-art baseline knowledge graph embedding models.
小貼士
登錄享
相關主題
注冊地址： 北京市海淀區羊坊店路18號2幢3層301-191