Xinyu Gao

Jiaxing Zhang, Ruyi Gan, Junjie Wang et al.

Nowadays, foundation models become one of fundamental infrastructures in artificial intelligence, paving ways to the general intelligence. However, the reality presents two urgent challenges: existing foundation models are dominated by the English-language community; users are often given limited resources and thus cannot always use foundation models. To support the development of the Chinese-language community, we introduce an open-source project, called Fengshenbang, which leads by the research center for Cognitive Computing and Natural Language (CCNL). Our project has comprehensive capabilities, including large pre-trained models, user-friendly APIs, benchmarks, datasets, and others. We wrap all these in three sub-projects: the Fengshenbang Model, the Fengshen Framework, and the Fengshen Benchmark. An open-source roadmap, Fengshenbang, aims to re-evaluate the open-source community of Chinese pre-trained large-scale models, prompting the development of the entire Chinese large-scale model community. We also want to build a user-centered open-source ecosystem to allow individuals to access the desired models to match their computing resources. Furthermore, we invite companies, colleges, and research institutions to collaborate with us to build the large-scale open-source model-based ecosystem. We hope that this project will be the foundation of Chinese cognitive intelligence.

Ping Yang, Junjie Wang, Ruyi Gan et al.

We propose a new paradigm for zero-shot learners that is format agnostic, i.e., it is compatible with any format and applicable to a list of language tasks, such as text classification, commonsense reasoning, coreference resolution, and sentiment analysis. Zero-shot learning aims to train a model on a given task such that it can address new learning tasks without any additional training. Our approach converts zero-shot learning into multiple-choice tasks, avoiding problems in commonly used large-scale generative models such as FLAN. It not only adds generalization ability to models but also significantly reduces the number of parameters. Our method shares the merits of efficient training and deployment. Our approach shows state-of-the-art performance on several benchmarks and produces satisfactory results on tasks such as natural language inference and text classification. Our model achieves this success with only 235M parameters, which is substantially smaller than state-of-the-art models with billions of parameters. The code and pre-trained models are available at https://github.com/IDEA-CCNL/Fengshenbang-LM .

Xinyu Gao, Xidong Mu, Wenqiang Yi et al.

A novel reconfigurable intelligent surface-aided multi-robot network is proposed, where multiple mobile robots are served by an access point (AP) through non-orthogonal multiple access (NOMA). The goal is to maximize the sum-rate of whole trajectories for the multi-robot system by jointly optimizing trajectories and NOMA decoding orders of robots, phase-shift coefficients of the RIS, and the power allocation of the AP, subject to predicted initial and final positions of robots and the quality of service (QoS) of each robot. To tackle this problem, an integrated machine learning (ML) scheme is proposed, which combines long short-term memory (LSTM)-autoregressive integrated moving average (ARIMA) model and dueling double deep Q-network (D$^{3}$QN) algorithm. For initial and final position prediction for robots, the LSTM-ARIMA is able to overcome the problem of gradient vanishment of non-stationary and non-linear sequences of data. For jointly determining the phase shift matrix and robots' trajectories, D$^{3}$QN is invoked for solving the problem of action value overestimation. Based on the proposed scheme, each robot holds an optimal trajectory based on the maximum sum-rate of a whole trajectory, which reveals that robots pursue long-term benefits for whole trajectory design. Numerical results demonstrated that: 1) LSTM-ARIMA model provides high accuracy predicting model; 2) The proposed D$^{3}$QN algorithm can achieve fast average convergence; and 3) RIS-NOMA networks have superior network performance compared to RIS-aided orthogonal counterparts.

Junyu Lu, Dixiang Zhang, Xiaojun Wu et al.

Recent advancements enlarge the capabilities of large language models (LLMs) in zero-shot image-to-text generation and understanding by integrating multi-modal inputs. However, such success is typically limited to English scenarios due to the lack of large-scale and high-quality non-English multi-modal resources, making it extremely difficult to establish competitive counterparts in other languages. In this paper, we introduce the Ziya-Visual series, a set of bilingual large-scale vision-language models (LVLMs) designed to incorporate visual semantics into LLM for multi-modal dialogue. Composed of Ziya-Visual-Base and Ziya-Visual-Chat, our models adopt the Querying Transformer from BLIP-2, further exploring the assistance of optimization schemes such as instruction tuning, multi-stage training and low-rank adaptation module for visual-language alignment. In addition, we stimulate the understanding ability of GPT-4 in multi-modal scenarios, translating our gathered English image-text datasets into Chinese and generating instruction-response through the in-context learning method. The experiment results demonstrate that compared to the existing LVLMs, Ziya-Visual achieves competitive performance across a wide range of English-only tasks including zero-shot image-text retrieval, image captioning, and visual question answering. The evaluation leaderboard accessed by GPT-4 also indicates that our models possess satisfactory image-text understanding and generation capabilities in Chinese multi-modal scenario dialogues. Code, demo and models are available at ~\url{https://huggingface.co/IDEA-CCNL/Ziya-BLIP2-14B-Visual-v1}.

Xinyu Gao, Wenqiang Yi, Yuanwei Liu et al.

Simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs) is a promising passive device that contributes to a full-space coverage via transmitting and reflecting the incident signal simultaneously. As a new paradigm in wireless communications, how to analyze the coverage and capacity performance of STAR-RISs becomes essential but challenging. To solve the coverage and capacity optimization (CCO) problem in STAR-RIS assisted networks, a multi-objective proximal policy optimization (MO-PPO) algorithm is proposed to handle long-term benefits than conventional optimization algorithms. To strike a balance between each objective, the MO-PPO algorithm provides a set of optimal solutions to form a Pareto front (PF), where any solution on the PF is regarded as an optimal result. Moreover, in order to improve the performance of the MO-PPO algorithm, two update strategies, i.e., action-value-based update strategy (AVUS) and loss function-based update strategy (LFUS), are investigated. For the AVUS, the improved point is to integrate the action values of both coverage and capacity and then update the loss function. For the LFUS, the improved point is only to assign dynamic weights for both loss functions of coverage and capacity, while the weights are calculated by a min-norm solver at every update. The numerical results demonstrated that the investigated update strategies outperform the fixed weights MO optimization algorithms in different cases, which includes a different number of sample grids, the number of STAR-RISs, the number of elements in the STAR-RISs, and the size of STAR-RISs. Additionally, the STAR-RIS assisted networks achieve better performance than conventional wireless networks without STAR-RISs. Moreover, with the same bandwidth, millimeter wave is able to provide higher capacity than sub-6 GHz, but at a cost of smaller coverage.

Xinyu Gao, Wenqiang Yi, Alexandros Agapitos et al.

Coverage and capacity are the important metrics for performance evaluation in wireless networks, while the coverage and capacity have several conflicting relationships, e.g. high transmit power contributes to large coverage but high inter-cell interference reduces the capacity performance. Therefore, in order to strike a balance between the coverage and capacity, a novel model is proposed for the coverage and capacity optimization of simultaneously transmitting and reflecting reconfigurable intelligent surfaces (STAR-RISs) assisted networks. To solve the coverage and capacity optimization (CCO) problem, a machine learning-based multi-objective optimization algorithm, i.e., the multi-objective proximal policy optimization (MO-PPO) algorithm, is proposed. In this algorithm, a loss function-based update strategy is the core point, which is able to calculate weights for both loss functions of coverage and capacity by a min-norm solver at each update. The numerical results demonstrate that the investigated update strategy outperforms the fixed weight-based MO algorithms.

Ting Han, Kunhao Pan, Xinyu Chen et al.

Bidirectional Encoder Representations from Transformers or BERT~\cite{devlin-etal-2019-bert} has been one of the base models for various NLP tasks due to its remarkable performance. Variants customized for different languages and tasks are proposed to further improve the performance. In this work, we investigate supervised continued pre-training~\cite{gururangan-etal-2020-dont} on BERT for Chinese topic classification task. Specifically, we incorporate prompt-based learning and contrastive learning into the pre-training. To adapt to the task of Chinese topic classification, we collect around 2.1M Chinese data spanning various topics. The pre-trained Chinese Topic Classification BERTs (TCBERTs) with different parameter sizes are open-sourced at \url{https://huggingface.co/IDEA-CCNL}.

Xinyu Gao, Ziyi Yang, Bingchen Gong et al.

Using parts of existing models to rebuild new models, commonly termed as example-based modeling, is a classical methodology in the realm of computer graphics. Previous works mostly focus on shape composition, making them very hard to use for realistic composition of 3D objects captured from real-world scenes. This leads to combining multiple NeRFs into a single 3D scene to achieve seamless appearance blending. However, the current SeamlessNeRF method struggles to achieve interactive editing and harmonious stitching for real-world scenes due to its gradient-based strategy and grid-based representation. To this end, we present an example-based modeling method that combines multiple Gaussian fields in a point-based representation using sample-guided synthesis. Specifically, as for composition, we create a GUI to segment and transform multiple fields in real time, easily obtaining a semantically meaningful composition of models represented by 3D Gaussian Splatting (3DGS). For texture blending, due to the discrete and irregular nature of 3DGS, straightforwardly applying gradient propagation as SeamlssNeRF is not supported. Thus, a novel sampling-based cloning method is proposed to harmonize the blending while preserving the original rich texture and content. Our workflow consists of three steps: 1) real-time segmentation and transformation of a Gaussian model using a well-tailored GUI, 2) KNN analysis to identify boundary points in the intersecting area between the source and target models, and 3) two-phase optimization of the target model using sampling-based cloning and gradient constraints. Extensive experimental results validate that our approach significantly outperforms previous works in terms of realistic synthesis, demonstrating its practicality. More demos are available at https://ingra14m.github.io/gs_stitching_website.

Ziyi Yang, Xinyu Gao, Yangtian Sun et al.

The recent advancements in 3D Gaussian splatting (3D-GS) have not only facilitated real-time rendering through modern GPU rasterization pipelines but have also attained state-of-the-art rendering quality. Nevertheless, despite its exceptional rendering quality and performance on standard datasets, 3D-GS frequently encounters difficulties in accurately modeling specular and anisotropic components. This issue stems from the limited ability of spherical harmonics (SH) to represent high-frequency information. To overcome this challenge, we introduce Spec-Gaussian, an approach that utilizes an anisotropic spherical Gaussian (ASG) appearance field instead of SH for modeling the view-dependent appearance of each 3D Gaussian. Additionally, we have developed a coarse-to-fine training strategy to improve learning efficiency and eliminate floaters caused by overfitting in real-world scenes. Our experimental results demonstrate that our method surpasses existing approaches in terms of rendering quality. Thanks to ASG, we have significantly improved the ability of 3D-GS to model scenes with specular and anisotropic components without increasing the number of 3D Gaussians. This improvement extends the applicability of 3D GS to handle intricate scenarios with specular and anisotropic surfaces. Project page is https://ingra14m.github.io/Spec-Gaussian-website/.

Junseo Kim, Guido Dumont, Xinyu Gao et al.

Dense 3D semantic occupancy perception is critical for mobile robots operating in pedestrian-rich environments, yet it remains underexplored compared to its application in autonomous driving. To address this gap, we present MobileOcc, a semantic occupancy dataset for mobile robots operating in crowded human environments. Our dataset is built using an annotation pipeline that incorporates static object occupancy annotations and a novel mesh optimization framework explicitly designed for human occupancy modeling. It reconstructs deformable human geometry from 2D images and subsequently refines and optimizes it using associated LiDAR point data. Using MobileOcc, we establish benchmarks for two tasks, i) Occupancy prediction and ii) Pedestrian velocity prediction, using different methods including monocular, stereo, and panoptic occupancy, with metrics and baseline implementations for reproducible comparison. Beyond occupancy prediction, we further assess our annotation method on 3D human pose estimation datasets. Results demonstrate that our method exhibits robust performance across different datasets.

Yuanwei Liu, Xiao Liu, Xinyu Gao et al.

The ongoing surge in applications of robotics brings both opportunities and challenges for the fifth-generation (5G) and beyond (B5G) of communication networks. This article focuses on 5G/B5G-enabled terrestrial robotic communications with an emphasis on distinct characteristics of such communications. Firstly, signal and spatial modeling for robotic communications are presented. To elaborate further, both the benefits and challenges derived from robots' mobility are discussed. As a further advance, a novel simultaneous localization and radio mapping (SLARM) framework is proposed for integrating localization and communications into robotic networks. Furthermore, dynamic trajectory design and resource allocation for both indoor and outdoor robots are provided to verify the performance of robotic communications in the context of typical robotic application scenarios.

Xinyu Gao, Yuanwei Liu, Xidong Mu

A novel simultaneous localization and radio mapping (SLARM) framework for communication-aware connected robots in the unknown indoor environment is proposed, where the simultaneous localization and mapping (SLAM) algorithm and the global geographic map recovery (GGMR) algorithm are leveraged to simultaneously construct a geographic map and a radio map named a channel power gain map. Specifically, the geographic map contains the information of a precise layout of obstacles and passable regions, and the radio map characterizes the position-dependent maximum expected channel power gain between the access point and the connected robot. Numerical results show that: 1) The pre-defined resolution in the SLAM algorithm and the proposed GGMR algorithm significantly affect the accuracy of the constructed radio map; and 2) The accuracy of radio map constructed by the SLARM framework is more than 78.78% when the resolution value smaller than 0.15m, and the accuracy reaches 91.95% when the resolution value is pre-defined as 0.05m.

Xinyu Gao, Yuanwei Liu, Xidong Mu

A novel intelligent reflecting surface (IRS)-aided multi-robot network is proposed, where multiple mobile wheeled robots are served by an access point (AP) through non-orthogonal multiple access (NOMA). The goal is to maximize the sum-rate of all robots by jointly optimizing trajectories and NOMA decoding orders of robots, reflecting coefficients of the IRS, and the power allocation of the AP, subject to the quality of service (QoS) of each robot. To tackle this problem, a dueling double deep Q-network (D^{3}QN) based algorithm is invoked for jointly determining the phase shift matrix and robots' trajectories. Specifically, the trajectories for robots contain a set of local optimal positions, which reveals that robots make the optimal decision at each step. Numerical results demonstrated that the proposed D^{3}QN algorithm outperforms the conventional algorithm, while the performance of IRS-NOMA network is better than the orthogonal multiple access (OMA) network.