分享

Recent years have witnessed a trend that large language model (LLM) based text-to-speech (TTS) emerges into the mainstream due to their high naturalness and zero-shot capacity. In this paradigm, speech signals are discretized into token sequences, which are modeled by an LLM with text as prompts and reconstructed by a token-based vocoder to waveforms. Obviously, speech tokens play a critical role in LLM-based TTS models. Current speech tokens are learned in an unsupervised manner, which lacks explicit semantic information and alignment to the text. In this paper, we propose to represent speech with supervised semantic tokens, which are derived from a multilingual speech recognition model by inserting vector quantization into the encoder. Based on the tokens, we further propose a scalable zero-shot TTS synthesizer, CosyVoice, which consists of an LLM for text-to-token generation and a conditional flow matching model for token-to-speech synthesis. Experimental results show that supervised semantic tokens significantly outperform existing unsupervised tokens in terms of content consistency and speaker similarity for zero-shot voice cloning. Moreover, we find that utilizing large-scale data further improves the synthesis performance, indicating the scalable capacity of CosyVoice. To the best of our knowledge, this is the first attempt to involve supervised speech tokens into TTS models.

Crowd Motion Generation is essential in entertainment industries such as animation and games as well as in strategic fields like urban simulation and planning. This new task requires an intricate integration of control and generation to realistically synthesize crowd dynamics under specific spatial and semantic constraints, whose challenges are yet to be fully explored. On the one hand, existing human motion generation models typically focus on individual behaviors, neglecting the complexities of collective behaviors. On the other hand, recent methods for multi-person motion generation depend heavily on pre-defined scenarios and are limited to a fixed, small number of inter-person interactions, thus hampering their practicality. To overcome these challenges, we introduce CrowdMoGen, a zero-shot text-driven framework that harnesses the power of Large Language Model (LLM) to incorporate the collective intelligence into the motion generation framework as guidance, thereby enabling generalizable planning and generation of crowd motions without paired training data. Our framework consists of two key components: 1) Crowd Scene Planner that learns to coordinate motions and dynamics according to specific scene contexts or introduced perturbations, and 2) Collective Motion Generator that efficiently synthesizes the required collective motions based on the holistic plans. Extensive quantitative and qualitative experiments have validated the effectiveness of our framework, which not only fills a critical gap by providing scalable and generalizable solutions for Crowd Motion Generation task but also achieves high levels of realism and flexibility.

Despite the success achieved by existing image generation and editing methods, current models still struggle with complex problems including intricate text prompts, and the absence of verification and self-correction mechanisms makes the generated images unreliable. Meanwhile, a single model tends to specialize in particular tasks and possess the corresponding capabilities, making it inadequate for fulfilling all user requirements. We propose GenArtist, a unified image generation and editing system, coordinated by a multimodal large language model (MLLM) agent. We integrate a comprehensive range of existing models into the tool library and utilize the agent for tool selection and execution. For a complex problem, the MLLM agent decomposes it into simpler sub-problems and constructs a tree structure to systematically plan the procedure of generation, editing, and self-correction with step-by-step verification. By automatically generating missing position-related inputs and incorporating position information, the appropriate tool can be effectively employed to address each sub-problem. Experiments demonstrate that GenArtist can perform various generation and editing tasks, achieving state-of-the-art performance and surpassing existing models such as SDXL and DALL-E 3, as can be seen in Fig. 1. Project page is https://zhenyuw16.github.io/GenArtist_page.

Although recent efforts have extended Neural Radiance Fields (NeRF) into LiDAR point cloud synthesis, the majority of existing works exhibit a strong dependence on precomputed poses. However, point cloud registration methods struggle to achieve precise global pose estimation, whereas previous pose-free NeRFs overlook geometric consistency in global reconstruction. In light of this, we explore the geometric insights of point clouds, which provide explicit registration priors for reconstruction. Based on this, we propose Geometry guided Neural LiDAR Fields(GeoNLF), a hybrid framework performing alternately global neural reconstruction and pure geometric pose optimization. Furthermore, NeRFs tend to overfit individual frames and easily get stuck in local minima under sparse-view inputs. To tackle this issue, we develop a selective-reweighting strategy and introduce geometric constraints for robust optimization. Extensive experiments on NuScenes and KITTI-360 datasets demonstrate the superiority of GeoNLF in both novel view synthesis and multi-view registration of low-frequency large-scale point clouds.

Traditional supervised 3D medical image segmentation models need voxel-level annotations, which require huge human effort, time, and cost. Semi-supervised learning (SSL) addresses this limitation of supervised learning by facilitating learning with a limited annotated and larger amount of unannotated training samples. However, state-of-the-art SSL models still struggle to fully exploit the potential of learning from unannotated samples. To facilitate effective learning from unannotated data, we introduce LLM-SegNet, which exploits a large language model (LLM) to integrate task-specific knowledge into our co-training framework. This knowledge aids the model in comprehensively understanding the features of the region of interest (ROI), ultimately leading to more efficient segmentation. Additionally, to further reduce erroneous segmentation, we propose a Unified Segmentation loss function. This loss function reduces erroneous segmentation by not only prioritizing regions where the model is confident in predicting between foreground or background pixels but also effectively addressing areas where the model lacks high confidence in predictions. Experiments on publicly available Left Atrium, Pancreas-CT, and Brats-19 datasets demonstrate the superior performance of LLM-SegNet compared to the state-of-the-art. Furthermore, we conducted several ablation studies to demonstrate the effectiveness of various modules and loss functions leveraged by LLM-SegNet.

In this paper, we propose to reformulate the blind image deblurring task to directly learn an inverse of the degradation model using a deep linear network. We introduce Deep Identity Learning (DIL), a novel learning strategy that includes a dedicated regularization term based on the properties of linear systems, to exploit the identity relation between the degradation and inverse degradation models. The salient aspect of our proposed framework is it neither relies on a deblurring dataset nor a single input blurred image (like Polyblur, a self-supervised method). Since it is purely image-data-independent, we term our model as Null-Shot deblurring Using Deep Identity Learning (NSD-DIL). We also provide an explicit representation of the learned deep linear network in a matrix form, called Deep Restoration Kernel (DRK) for deblurring task. The proposed framework detours the typical degradation kernel estimation step involved in most of the existing blind deblurring solutions by the proposition of our Random Kernel Gallery (RKG) dataset. In this work, we focus on the restoration of mild blur images, generated by small out-of-focus, lens blur, or slight camera motion, which often occurs in real images. Our experiments show that the proposed method outperforms both traditional and deep learning based deblurring methods, with at least an order of 100 lesser computational resources. The proposed NSD-DIL method can be effortlessly extended to the Image Super-Resolution (ISR) task as well to restore the low-resolution images with fine details. The NSD-DIL model and its kernel form representation (DRK) are lightweight yet robust and restore the mild blur input in a fraction of a second. Hence, more suitable for wide real-time applications.

Cross-lingual speech emotion recognition (SER) is important for a wide range of everyday applications. While recent SER research relies heavily on large pretrained models for emotion training, existing studies often concentrate solely on the final transformer layer of these models. However, given the task-specific nature and hierarchical architecture of these models, each transformer layer encapsulates different levels of information. Leveraging this hierarchical structure, our study focuses on the information embedded across different layers. Through an examination of layer feature similarity across different languages, we propose a novel strategy called a layer-anchoring mechanism to facilitate emotion transfer in cross-lingual SER tasks. Our approach is evaluated using two distinct language affective corpora (MSP-Podcast and BIIC-Podcast), achieving a best UAR performance of 60.21% on the BIIC-podcast corpus. The analysis uncovers interesting insights into the behavior of popular pretrained models.

This paper introduces BI-Directional DEliberation Reasoning (BIDDER), a novel reasoning approach to enhance the decision rationality of language models. Traditional reasoning methods typically rely on historical information and employ uni-directional (left-to-right) reasoning strategy. This lack of bi-directional deliberation reasoning results in limited awareness of potential future outcomes and insufficient integration of historical context, leading to suboptimal decisions. BIDDER addresses this gap by incorporating principles of rational decision-making, specifically managing uncertainty and predicting expected utility. Our approach involves three key processes: Inferring hidden states to represent uncertain information in the decision-making process from historical data; Using these hidden states to predict future potential states and potential outcomes; Integrating historical information (past contexts) and long-term outcomes (future contexts) to inform reasoning. By leveraging bi-directional reasoning, BIDDER ensures thorough exploration of both past and future contexts, leading to more informed and rational decisions. We tested BIDDER's effectiveness in two well-defined scenarios: Poker (Limit Texas Hold'em) and Negotiation. Our experiments demonstrate that BIDDER significantly improves the decision-making capabilities of LLMs and LLM agents.

Computer programming initially required humans to directly translate their goals into machine code. These goals could have easily been expressed as a written (or human) language directive. Computers, however, had no capacity to satisfactorily interpret written language. Large language model's provide exactly this capability; automatic generation of computer programs or even assembly code from written language directives. This research examines dynamic code execution of written language directives within the context of a running application. It implements a text editor whose business logic is purely backed by large language model prompts. That is, the program's execution uses prompts and written language directives to dynamically generate application logic at the point in time it is needed. The research clearly shows how written language directives, backed by a large language model, offer radically new programming and operating system paradigms. For example, empowerment of users to directly implement requirements via written language directives, thus supplanting the need for a team ofprogrammers, a release schedule and the like. Or, new security mechanisms where static executables, always a target for reverse engineering or fuzzing, no longer exist. They are replaced by ephemeral executables that may continually change, be completely removed, and are easily updated.

Few-shot Named Entity Recognition (NER), the task of identifying named entities with only a limited amount of labeled data, has gained increasing significance in natural language processing. While existing methodologies have shown some effectiveness, such as enriching label semantics through various prompting modes or employing metric learning techniques, their performance exhibits limited robustness across diverse domains due to the lack of rich knowledge in their pre-trained models. To address this issue, we propose CLLMFS, a Contrastive Learning enhanced Large Language Model (LLM) Framework for Few-Shot Named Entity Recognition, achieving promising results with limited training data. Considering the impact of LLM's internal representations on downstream tasks, CLLMFS integrates Low-Rank Adaptation (LoRA) and contrastive learning mechanisms specifically tailored for few-shot NER. By enhancing the model's internal representations, CLLMFS effectively improves both entity boundary awareness ability and entity recognition accuracy. Our method has achieved state-of-the-art performance improvements on F1-score ranging from 2.58\% to 97.74\% over existing best-performing methods across several recognized benchmarks. Furthermore, through cross-domain NER experiments conducted on multiple datasets, we have further validated the robust generalization capability of our method. Our code will be released in the near future.

The current public datasets for speech recognition (ASR) tend not to focus specifically on the fairness aspect, such as performance across different demographic groups. This paper introduces a novel dataset, Fair-Speech, a publicly released corpus to help researchers evaluate their ASR models for accuracy across a diverse set of self-reported demographic information, such as age, gender, ethnicity, geographic variation and whether the participants consider themselves native English speakers. Our dataset includes approximately 26.5K utterances in recorded speech by 593 people in the United States, who were paid to record and submit audios of themselves saying voice commands. We also provide ASR baselines, including on models trained on transcribed and untranscribed social media videos and open source models.

In this paper, we study the problem of estimating the covariance matrix under differential privacy, where the underlying covariance matrix is assumed to be sparse and of high dimensions. We propose a new method, called DP-Thresholding, to achieve a non-trivial $\ell_2$-norm based error bound, which is significantly better than the existing ones from adding noise directly to the empirical covariance matrix. We also extend the $\ell_2$-norm based error bound to a general $\ell_w$-norm based one for any $1\leq w\leq \infty$, and show that they share the same upper bound asymptotically. Our approach can be easily extended to local differential privacy. Experiments on the synthetic datasets show consistent results with our theoretical claims.