分享

Large Language Model (LLM) based text-to-speech (TTS) systems have demonstrated remarkable capabilities in handling large speech datasets and generating natural speech for new speakers. However, LLM-based TTS models are not robust as the generated output can contain repeating words, missing words and mis-aligned speech (referred to as hallucinations or attention errors), especially when the text contains multiple occurrences of the same token. We examine these challenges in an encoder-decoder transformer model and find that certain cross-attention heads in such models implicitly learn the text and speech alignment when trained for predicting speech tokens for a given text. To make the alignment more robust, we propose techniques utilizing CTC loss and attention priors that encourage monotonic cross-attention over the text tokens. Our guided attention training technique does not introduce any new learnable parameters and significantly improves robustness of LLM-based TTS models.

Dental panoramic radiographs offer vast diagnostic opportunities, but training supervised deep learning networks for automatic analysis of those radiology images is hampered by a shortage of labeled data. Here, a different perspective on this problem is introduced. A semi-supervised learning framework is proposed to classify thirteen dental conditions on panoramic radiographs, with a particular emphasis on teeth. Large language models were explored to annotate the most common dental conditions based on dental reports. Additionally, a masked autoencoder was employed to pre-train the classification neural network, and a Vision Transformer was used to leverage the unlabeled data. The analyses were validated using two of the most extensive datasets in the literature, comprising 8,795 panoramic radiographs and 8,029 paired reports and images. Encouragingly, the results consistently met or surpassed the baseline metrics for the Matthews correlation coefficient. A comparison of the proposed solution with human practitioners, supported by statistical analysis, highlighted its effectiveness and performance limitations; based on the degree of agreement among specialists, the solution demonstrated an accuracy level comparable to that of a junior specialist.

Intelligent agents need to autonomously navigate and interact within contextual environments to perform a wide range of daily tasks based on human-level instructions. These agents require a foundational understanding of the world, incorporating common sense and knowledge, to interpret such instructions. Moreover, they must possess precise low-level skills for movement and interaction to execute the detailed task plans derived from these instructions. In this work, we address the task of synthesizing continuous human-object interactions for manipulating large objects within contextual environments, guided by human-level instructions. Our goal is to generate synchronized object motion, full-body human motion, and detailed finger motion, all essential for realistic interactions. Our framework consists of a large language model (LLM) planning module and a low-level motion generator. We use LLMs to deduce spatial object relationships and devise a method for accurately determining their positions and orientations in target scene layouts. Additionally, the LLM planner outlines a detailed task plan specifying a sequence of sub-tasks. This task plan, along with the target object poses, serves as input for our low-level motion generator, which seamlessly alternates between navigation and interaction modules. We present the first complete system that can synthesize object motion, full-body motion, and finger motion simultaneously from human-level instructions. Our experiments demonstrate the effectiveness of our high-level planner in generating plausible target layouts and our low-level motion generator in synthesizing realistic interactions for diverse objects. Please refer to our project page for more results: https://hoifhli.github.io/.

In this work, we propose a Switch-Conformer-based MoE system named SC-MoE for unified streaming and non-streaming code-switching (CS) automatic speech recognition (ASR), where we design a streaming MoE layer consisting of three language experts, which correspond to Mandarin, English, and blank, respectively, and equipped with a language identification (LID) network with a Connectionist Temporal Classification (CTC) loss as a router in the encoder of SC-MoE to achieve a real-time streaming CS ASR system. To further utilize the language information embedded in text, we also incorporate MoE layers into the decoder of SC-MoE. In addition, we introduce routers into every MoE layer of the encoder and the decoder and achieve better recognition performance. Experimental results show that the SC-MoE significantly improves CS ASR performances over baseline with comparable computational efficiency.

Sequential recommendation systems fundamentally rely on users' historical interaction sequences, which are often contaminated by noisy interactions. Identifying these noisy interactions accurately without additional information is particularly difficult due to the lack of explicit supervisory signals to denote noise. Large Language Models (LLMs), equipped with extensive open knowledge and semantic reasoning abilities, present a promising avenue to bridge this information gap. However, employing LLMs for denoising in sequential recommendation introduces notable challenges: 1) Direct application of pretrained LLMs may not be competent for the denoising task, frequently generating nonsensical responses; 2) Even after fine-tuning, the reliability of LLM outputs remains questionable, especially given the complexity of the task and th inherent hallucinatory issue of LLMs. To tackle these challenges, we propose LLM4DSR, a tailored approach for denoising sequential recommendation using LLMs. We constructed a self-supervised fine-tuning task to activate LLMs' capabilities to identify noisy items and suggest replacements. Furthermore, we developed an uncertainty estimation module that ensures only high-confidence responses are utilized for sequence corrections. Remarkably, LLM4DSR is model-agnostic, allowing the corrected sequences to be flexibly applied across various recommendation models. Extensive experiments validate the superiority of LLM4DSR over existing methods across three datasets and three recommendation backbones.

In the realm of event prediction, temporal knowledge graph forecasting (TKGF) stands as a pivotal technique. Previous approaches face the challenges of not utilizing experience during testing and relying on a single short-term history, which limits adaptation to evolving data. In this paper, we introduce the Online Neural-Symbolic Event Prediction (ONSEP) framework, which innovates by integrating dynamic causal rule mining (DCRM) and dual history augmented generation (DHAG). DCRM dynamically constructs causal rules from real-time data, allowing for swift adaptation to new causal relationships. In parallel, DHAG merges short-term and long-term historical contexts, leveraging a bi-branch approach to enrich event prediction. Our framework demonstrates notable performance enhancements across diverse datasets, with significant Hit@k (k=1,3,10) improvements, showcasing its ability to augment large language models (LLMs) for event prediction without necessitating extensive retraining. The ONSEP framework not only advances the field of TKGF but also underscores the potential of neural-symbolic approaches in adapting to dynamic data environments.

Human motion prediction is crucial for human-centric multimedia understanding and interacting. Current methods typically rely on ground truth human poses as observed input, which is not practical for real-world scenarios where only raw visual sensor data is available. To implement these methods in practice, a pre-phrase of pose estimation is essential. However, such two-stage approaches often lead to performance degradation due to the accumulation of errors. Moreover, reducing raw visual data to sparse keypoint representations significantly diminishes the density of information, resulting in the loss of fine-grained features. In this paper, we propose \textit{LiDAR-HMP}, the first single-LiDAR-based 3D human motion prediction approach, which receives the raw LiDAR point cloud as input and forecasts future 3D human poses directly. Building upon our novel structure-aware body feature descriptor, LiDAR-HMP adaptively maps the observed motion manifold to future poses and effectively models the spatial-temporal correlations of human motions for further refinement of prediction results. Extensive experiments show that our method achieves state-of-the-art performance on two public benchmarks and demonstrates remarkable robustness and efficacy in real-world deployments.

In this study, we aim to develop a domestic service robot (DSR) that, guided by open-vocabulary instructions, can carry everyday objects to the specified pieces of furniture. Few existing methods handle mobile manipulation tasks with open-vocabulary instructions in the image retrieval setting, and most do not identify both the target objects and the receptacles. We propose the Dual-Mode Multimodal Ranking model (DM2RM), which enables images of both the target objects and receptacles to be retrieved using a single model based on multimodal foundation models. We introduce a switching mechanism that leverages a mode token and phrase identification via a large language model to switch the embedding space based on the prediction target. To evaluate the DM2RM, we construct a novel dataset including real-world images collected from hundreds of building-scale environments and crowd-sourced instructions with referring expressions. The evaluation results show that the proposed DM2RM outperforms previous approaches in terms of standard metrics in image retrieval settings. Furthermore, we demonstrate the application of the DM2RM on a standardized real-world DSR platform including fetch-and-carry actions, where it achieves a task success rate of 82% despite the zero-shot transfer setting. Demonstration videos, code, and more materials are available at https://kkrr10.github.io/dm2rm/.

Large Transformer networks are increasingly used in settings where low inference latency can improve the end-user experience and enable new applications. However, autoregressive inference is resource intensive and requires parallelism for efficiency. Parallelism introduces collective communication that is both expensive and represents a phase when hardware resources are underutilized. Towards mitigating this, Kraken is an evolution of the standard Transformer architecture that is designed to complement existing tensor parallelism schemes for efficient inference on multi-device systems. By introducing a fixed degree of intra-layer model parallelism, the architecture allows collective operations to be overlapped with compute, decreasing latency and increasing hardware utilization. When trained on OpenWebText, Kraken models reach a similar perplexity as standard Transformers while also preserving their language modeling capabilities when evaluated on the SuperGLUE benchmark. Importantly, when tested on multi-GPU systems using TensorRT-LLM engines, Kraken speeds up Time To First Token by a mean of 35.6% across a range of model sizes, context lengths, and degrees of tensor parallelism.

We develop a large language model (LLM) based automatic speech recognition (ASR) system that can be contextualized by providing keywords as prior information in text prompts. We adopt decoder-only architecture and use our in-house LLM, PLaMo-100B, pre-trained from scratch using datasets dominated by Japanese and English texts as the decoder. We adopt a pre-trained Whisper encoder as an audio encoder, and the audio embeddings from the audio encoder are projected to the text embedding space by an adapter layer and concatenated with text embeddings converted from text prompts to form inputs to the decoder. By providing keywords as prior information in the text prompts, we can contextualize our LLM-based ASR system without modifying the model architecture to transcribe ambiguous words in the input audio accurately. Experimental results demonstrate that providing keywords to the decoder can significantly improve the recognition performance of rare and ambiguous words.

Flavor development in the food industry is increasingly challenged by the need for rapid innovation and precise flavor profile creation. Traditional flavor research methods typically rely on iterative, subjective testing, which lacks the efficiency and scalability required for modern demands. This paper presents three contributions to address the challenges. Firstly, we define a new problem domain for scientific agents in flavor science, conceptualized as the generation of hypotheses for flavor profile sourcing and understanding. To facilitate research in this area, we introduce the FoodPuzzle, a challenging benchmark consisting of 978 food items and 1,766 flavor molecules profiles. We propose a novel Scientific Agent approach, integrating in-context learning and retrieval augmented techniques to generate grounded hypotheses in the domain of food science. Experimental results indicate that our model significantly surpasses traditional methods in flavor profile prediction tasks, demonstrating its potential to transform flavor development practices.

Parametric Computer-Aided Design (CAD) is central to contemporary mechanical design. However, it encounters challenges in achieving precise parametric sketch modeling and lacks practical evaluation metrics suitable for mechanical design. We harness the capabilities of pre-trained foundation models, renowned for their successes in natural language processing and computer vision, to develop generative models specifically for CAD. These models are adept at understanding complex geometries and design reasoning, a crucial advancement in CAD technology. In this paper, we propose CadVLM, an end-to-end vision language model for CAD generation. Our approach involves adapting pre-trained foundation models to manipulate engineering sketches effectively, integrating both sketch primitive sequences and sketch images. Extensive experiments demonstrate superior performance on multiple CAD sketch generation tasks such as CAD autocompletion, CAD autoconstraint, and image conditional generation. To our knowledge, this is the first instance of a multimodal Large Language Model (LLM) being successfully applied to parametric CAD generation, representing a pioneering step in the field of computer-aided mechanical design.