Sicong Liu

Team HY-World, Chenjie Cao, Xuhui Zuo et al.

We introduce HY-World 2.0, a multi-modal world model framework that advances our prior project HY-World 1.0. HY-World 2.0 accommodates diverse input modalities, including text prompts, single-view images, multi-view images, and videos, and produces 3D world representations. With text or single-view image inputs, the model performs world generation, synthesizing high-fidelity, navigable 3D Gaussian Splatting (3DGS) scenes. This is achieved through a four-stage method: a) Panorama Generation with HY-Pano 2.0, b) Trajectory Planning with WorldNav, c) World Expansion with WorldStereo 2.0, and d) World Composition with WorldMirror 2.0. Specifically, we introduce key innovations to enhance panorama fidelity, enable 3D scene understanding and planning, and upgrade WorldStereo, our keyframe-based view generation model with consistent memory. We also upgrade WorldMirror, a feed-forward model for universal 3D prediction, by refining model architecture and learning strategy, enabling world reconstruction from multi-view images or videos. Also, we introduce WorldLens, a high-performance 3DGS rendering platform featuring a flexible engine-agnostic architecture, automatic IBL lighting, efficient collision detection, and training-rendering co-design, enabling interactive exploration of 3D worlds with character support. Extensive experiments demonstrate that HY-World 2.0 achieves state-of-the-art performance on several benchmarks among open-source approaches, delivering results comparable to the closed-source model Marble. We release all model weights, code, and technical details to facilitate reproducibility and support further research on 3D world models.

Sicong Liu, Bin Guo, Cheng Fang et al.

The emerging field of artificial intelligence of things (AIoT, AI+IoT) is driven by the widespread use of intelligent infrastructures and the impressive success of deep learning (DL). With the deployment of DL on various intelligent infrastructures featuring rich sensors and weak DL computing capabilities, a diverse range of AIoT applications has become possible. However, DL models are notoriously resource-intensive. Existing research strives to realize near-/realtime inference of AIoT live data and low-cost training using AIoT datasets on resource-scare infrastructures. Accordingly, the accuracy and responsiveness of DL models are bounded by resource availability. To this end, the algorithm-system co-design that jointly optimizes the resource-friendly DL models and model-adaptive system scheduling improves the runtime resource availability and thus pushes the performance boundary set by the standalone level. Unlike previous surveys on resource-friendly DL models or hand-crafted DL compilers/frameworks with partially fine-tuned components, this survey aims to provide a broader optimization space for more free resource-performance tradeoffs. The cross-level optimization landscape involves various granularity, including the DL model, computation graph, operator, memory schedule, and hardware instructor in both on-device and distributed paradigms. Furthermore, due to the dynamic nature of AIoT context, which includes heterogeneous hardware, agnostic sensing data, varying user-specified performance demands, and resource constraints, this survey explores the context-aware inter-/intra-device controllers for automatic cross-level adaptation. Additionally, we identify some potential directions for resource-efficient AIoT systems. By consolidating problems and techniques scattered over diverse levels, we aim to help readers understand their connections and stimulate further discussions.

Xiao Tang, Sicong Liu, Xiaojiang Du et al.

Massive random access of devices in the emerging Open Radio Access Network (O-RAN) brings great challenge to the access control and management. Exploiting the bursting nature of the access requests, sparse active user detection (SAUD) is an efficient enabler towards efficient access management, but the sparsity might be deteriorated in case of uncoordinated massive access requests. To dynamically preserve the sparsity of access requests, a reinforcement-learning (RL)-assisted scheme of closed-loop access control utilizing the access class barring technique is proposed, where the RL policy is determined through continuous interaction between the RL agent, i.e., a next generation node base (gNB), and the environment. The proposed scheme can be implemented by the near-real-time RAN intelligent controller (near-RT RIC) in O-RAN, supporting rapid switching between heterogeneous vertical applications, such as mMTC and uRLLC services. Moreover, a data-driven scheme of deep-RL-assisted SAUD is proposed to resolve highly complex environments with continuous and high-dimensional state and action spaces, where a replay buffer is applied for automatic large-scale data collection. An actor-critic framework is formulated to incorporate the strategy-learning modules into the near-RT RIC. Simulation results show that the proposed schemes can achieve superior performance in both access efficiency and user detection accuracy over the benchmark scheme for different heterogeneous services with massive access requests.

Sicong Liu, Xiaochen Li, Zimu Zhou et al.

The ubiquity of camera-embedded devices and the advances in deep learning have stimulated various intelligent mobile video applications. These applications often demand on-device processing of video streams to deliver real-time, high-quality services for privacy and robustness concerns. However, the performance of these applications is constrained by the raw video streams, which tend to be taken with small-aperture cameras of ubiquitous mobile platforms in dim light. Despite extensive low-light video enhancement solutions, they are unfit for deployment to mobile devices due to their complex models and and ignorance of system dynamics like energy budgets. In this paper, we propose AdaEnlight, an energy-aware low-light video stream enhancement system on mobile devices. It achieves real-time video enhancement with competitive visual quality while allowing runtime behavior adaptation to the platform-imposed dynamic energy budgets. We report extensive experiments on diverse datasets, scenarios, and platforms and demonstrate the superiority of AdaEnlight compared with state-of-the-art low-light image and video enhancement solutions.

Younan Mou, Sicong Liu

The underwater propagation environment for visible light signals is affected by complex factors such as absorption, shadowing, and reflection, making it very challengeable to achieve effective underwater visible light communication (UVLC) channel estimation. It is difficult for the UVLC channel to be sparse represented in the time and frequency domains, which limits the chance of using sparse signal processing techniques to achieve better performance of channel estimation. To this end, a compressed sensing (CS) based framework is established in this paper by fully exploiting the sparsity of the underwater visible light channel in the distance domain of the propagation links. In order to solve the sparse recovery problem and achieve more accurate UVLC channel estimation, a sparse learning based underwater visible light channel estimation (SL-UVCE) scheme is proposed. Specifically, a deep-unfolding neural network mimicking the classical iterative sparse recovery algorithm of approximate message passing (AMP) is employed, which decomposes the iterations of AMP into a series of layers with different learnable parameters. Compared with the existing non-CS-based and CS-based schemes, the proposed scheme shows better performance of accuracy in channel estimation, especially in severe conditions such as insufficient measurement pilots and large number of multipath components.

Zibo Zhao, Zeqiang Lai, Qingxiang Lin et al.

We present Hunyuan3D 2.0, an advanced large-scale 3D synthesis system for generating high-resolution textured 3D assets. This system includes two foundation components: a large-scale shape generation model -- Hunyuan3D-DiT, and a large-scale texture synthesis model -- Hunyuan3D-Paint. The shape generative model, built on a scalable flow-based diffusion transformer, aims to create geometry that properly aligns with a given condition image, laying a solid foundation for downstream applications. The texture synthesis model, benefiting from strong geometric and diffusion priors, produces high-resolution and vibrant texture maps for either generated or hand-crafted meshes. Furthermore, we build Hunyuan3D-Studio -- a versatile, user-friendly production platform that simplifies the re-creation process of 3D assets. It allows both professional and amateur users to manipulate or even animate their meshes efficiently. We systematically evaluate our models, showing that Hunyuan3D 2.0 outperforms previous state-of-the-art models, including the open-source models and closed-source models in geometry details, condition alignment, texture quality, and etc. Hunyuan3D 2.0 is publicly released in order to fill the gaps in the open-source 3D community for large-scale foundation generative models. The code and pre-trained weights of our models are available at: https://github.com/Tencent/Hunyuan3D-2

Sicong Liu, Yanxian Huang, Mingwei Liu et al.

Code generation tasks aim to automate the conversion of user requirements into executable code, significantly reducing manual development efforts and enhancing software productivity. The emergence of large language models (LLMs) has significantly advanced code generation, though their efficiency is still impacted by certain inherent architectural constraints. Each token generation necessitates a complete inference pass, requiring persistent retention of contextual information in memory and escalating resource consumption. While existing research prioritizes inference-phase optimizations such as prompt compression and model quantization, the generation phase remains underexplored. To tackle these challenges, we propose a knowledge-infused framework named ShortCoder, which optimizes code generation efficiency while preserving semantic equivalence and readability. In particular, we introduce: (1) ten syntax-level simplification rules for Python, derived from AST-preserving transformations, achieving 18.1% token reduction without functional compromise; (2) a hybrid data synthesis pipeline integrating rule-based rewriting with LLM-guided refinement, producing ShorterCodeBench, a corpus of validated tuples of original code and simplified code with semantic consistency; (3) a fine-tuning strategy that injects conciseness awareness into the base LLMs. Extensive experimental results demonstrate that ShortCoder consistently outperforms state-of-the-art methods on HumanEval, achieving an improvement of 18.1%-37.8% in generation efficiency over previous methods while ensuring the performance of code generation.

Zijie Wu, Lixin Xu, Puhua Jiang et al.

Video-guided 3D animation holds immense potential for content creation, offering intuitive and precise control over dynamic assets. However, practical deployment faces a critical yet frequently overlooked hurdle: the pose misalignment dilemma. In real-world scenarios, the initial pose of a user-provided static mesh rarely aligns with the starting frame of a reference video. Naively forcing a mesh to follow a mismatched trajectory inevitably leads to severe geometric distortion or animation failure. To address this, we present Rectified Dynamic Mesh (R-DMesh), a unified framework designed to generate high-fidelity 4D meshes that are ``rectified'' to align with video context. Unlike standard motion transfer approaches, our method introduces a novel VAE that explicitly disentangles the input into a conditional base mesh, relative motion trajectories, and a crucial rectification jump offset. This offset is learned to automatically transform the arbitrary pose of the input mesh to match the video's initial state before animation begins. We process these components via a Triflow Attention mechanism, which leverages vertex-wise geometric features to modulate the three orthogonal flows, ensuring physical consistency and local rigidity during the rectification and animation process. For generation, we employ a Rectified Flow-based Diffusion Transformer conditioned on pre-trained video latents, effectively transferring rich spatio-temporal priors to the 3D domain. To support this task, we construct Video-RDMesh, a large-scale dataset of over 500k dynamic mesh sequences specifically curated to simulate pose misalignment. Extensive experiments demonstrate that R-DMesh not only solves the alignment problem but also enables robust downstream applications, including pose retargeting and holistic 4D generation.

Xiaochen Li, Sicong Liu, Bin Guo et al.

GB-scale large apps like on-device LLMs and rich media editors are becoming the next-generation trend, but their heavy memory and I/O demands, especially during multitasking, cause devices to reclaim or kill processes, turning warm apps into cold launches. The challenge lies not in storing them, but in fast, accurate launching. For users, 1s is the usability cliff, yet our measurements show 86.6\% of GB-scale cold launches exceed it. Also, Android Vitals flags only $\geq$ 5s as slow, exposing a large satisfaction gap. Existing optimizations are designed in isolation and conflict. For example, preloading reduces I/O stalls but consumes scarce memory and is undone by reclamation, while reclamation and killing free memory but sacrifice background survivability, leading to repeated cold relaunches. Our key insight is that, although multitasking makes runtime behavior complex, each app's file access pattern remains predictable. The challenge lies in exploiting this predictability, i.e., preloading without exhausting memory, reclaiming without undoing gains, and killing selectively to preserve background survivability. We introduce AppFlow, a prediction-based system-wide scheduler that integrates a Selective File Preloader, an Adaptive Memory Reclaimer, and a Context-Aware Process Killer. Implemented across the Android framework and Linux kernel without app changes, AppFlow cuts GB-scale cold-launch latency by 66.5\% (e.g., 2s$\rightarrow$690ms) and sustains 95\% of launches within 1s over a 100-day test, significantly improving responsiveness and multitasking experience.

Tiankuo Wei, Sicong Liu

Visible light positioning (VLP) has drawn plenty of attention as a promising indoor positioning technique. However, in nonstationary environments, the performance of VLP is limited because of the highly time-varying channels. To improve the positioning accuracy and generalization capability in nonstationary environments, a cooperative VLP scheme based on federated learning (FL) is proposed in this paper. Exploiting the FL framework, a global model adaptive to environmental changes can be jointly trained by users without sharing private data of users. Moreover, a Cooperative Visible-light Positioning Network (CVPosNet) is proposed to accelerate the convergence rate and improve the positioning accuracy. Simulation results show that the proposed scheme outperforms the benchmark schemes, especially in nonstationary environments.

Xiao Tang, Sicong Liu

Visible light communication (VLC) has been widely applied as a promising solution for modern short range communication. When it comes to the deployment of LED arrays in VLC networks, the emerging ultra-dense network (UDN) technology can be adopted to expand the VLC network's capacity. However, the problem of inter-cell interference (ICI) mitigation and efficient power control in the VLC-based UDN is still a critical challenge. To this end, a reinforcement learning (RL) based VLC UDN architecture is devised in this paper. The deployment of the cells is optimized via spatial reuse to mitigate ICI. An RL-based algorithm is proposed to dynamically optimize the policy of power and interference control, maximizing the system utility in the complicated and dynamic environment. Simulation results demonstrate the superiority of the proposed scheme, it increase the system utility and achievable data rate while reducing the energy consumption and ICI, which outperforms the benchmark scheme.

Lei Ding, Haitao Guo, Sicong Liu et al.

Semantic change detection (SCD) extends the multi-class change detection (MCD) task to provide not only the change locations but also the detailed land-cover/land-use (LCLU) categories before and after the observation intervals. This fine-grained semantic change information is very useful in many applications. Recent studies indicate that the SCD can be modeled through a triple-branch Convolutional Neural Network (CNN), which contains two temporal branches and a change branch. However, in this architecture, the communications between the temporal branches and the change branch are insufficient. To overcome the limitations in existing methods, we propose a novel CNN architecture for the SCD, where the semantic temporal features are merged in a deep CD unit. Furthermore, we elaborate on this architecture to reason the bi-temporal semantic correlations. The resulting Bi-temporal Semantic Reasoning Network (Bi-SRNet) contains two types of semantic reasoning blocks to reason both single-temporal and cross-temporal semantic correlations, as well as a novel loss function to improve the semantic consistency of change detection results. Experimental results on a benchmark dataset show that the proposed architecture obtains significant accuracy improvements over the existing approaches, while the added designs in the Bi-SRNet further improves the segmentation of both semantic categories and the changed areas. The codes in this paper are accessible at: github.com/ggsDing/Bi-SRNet.

Xianghui Yang, Huiwen Shi, Bowen Zhang et al.

While 3D generative models have greatly improved artists' workflows, the existing diffusion models for 3D generation suffer from slow generation and poor generalization. To address this issue, we propose a two-stage approach named Hunyuan3D 1.0 including a lite version and a standard version, that both support text- and image-conditioned generation. In the first stage, we employ a multi-view diffusion model that efficiently generates multi-view RGB in approximately 4 seconds. These multi-view images capture rich details of the 3D asset from different viewpoints, relaxing the tasks from single-view to multi-view reconstruction. In the second stage, we introduce a feed-forward reconstruction model that rapidly and faithfully reconstructs the 3D asset given the generated multi-view images in approximately 7 seconds. The reconstruction network learns to handle noises and in-consistency introduced by the multi-view diffusion and leverages the available information from the condition image to efficiently recover the 3D structure. Our framework involves the text-to-image model, i.e., Hunyuan-DiT, making it a unified framework to support both text- and image-conditioned 3D generation. Our standard version has 3x more parameters than our lite and other existing model. Our Hunyuan3D 1.0 achieves an impressive balance between speed and quality, significantly reducing generation time while maintaining the quality and diversity of the produced assets.

Team Hunyuan3D, Shuhui Yang, Mingxin Yang et al.

3D AI-generated content (AIGC) is a passionate field that has significantly accelerated the creation of 3D models in gaming, film, and design. Despite the development of several groundbreaking models that have revolutionized 3D generation, the field remains largely accessible only to researchers, developers, and designers due to the complexities involved in collecting, processing, and training 3D models. To address these challenges, we introduce Hunyuan3D 2.1 as a case study in this tutorial. This tutorial offers a comprehensive, step-by-step guide on processing 3D data, training a 3D generative model, and evaluating its performance using Hunyuan3D 2.1, an advanced system for producing high-resolution, textured 3D assets. The system comprises two core components: the Hunyuan3D-DiT for shape generation and the Hunyuan3D-Paint for texture synthesis. We will explore the entire workflow, including data preparation, model architecture, training strategies, evaluation metrics, and deployment. By the conclusion of this tutorial, you will have the knowledge to finetune or develop a robust 3D generative model suitable for applications in gaming, virtual reality, and industrial design.

Zeqiang Lai, Yunfei Zhao, Haolin Liu et al.

In this report, we present Hunyuan3D 2.5, a robust suite of 3D diffusion models aimed at generating high-fidelity and detailed textured 3D assets. Hunyuan3D 2.5 follows two-stages pipeline of its previous version Hunyuan3D 2.0, while demonstrating substantial advancements in both shape and texture generation. In terms of shape generation, we introduce a new shape foundation model -- LATTICE, which is trained with scaled high-quality datasets, model-size, and compute. Our largest model reaches 10B parameters and generates sharp and detailed 3D shape with precise image-3D following while keeping mesh surface clean and smooth, significantly closing the gap between generated and handcrafted 3D shapes. In terms of texture generation, it is upgraded with phyiscal-based rendering (PBR) via a novel multi-view architecture extended from Hunyuan3D 2.0 Paint model. Our extensive evaluation shows that Hunyuan3D 2.5 significantly outperforms previous methods in both shape and end-to-end texture generation.

HunyuanWorld Team, Zhenwei Wang, Yuhao Liu et al.

Creating immersive and playable 3D worlds from texts or images remains a fundamental challenge in computer vision and graphics. Existing world generation approaches typically fall into two categories: video-based methods that offer rich diversity but lack 3D consistency and rendering efficiency, and 3D-based methods that provide geometric consistency but struggle with limited training data and memory-inefficient representations. To address these limitations, we present HunyuanWorld 1.0, a novel framework that combines the best of both worlds for generating immersive, explorable, and interactive 3D scenes from text and image conditions. Our approach features three key advantages: 1) 360° immersive experiences via panoramic world proxies; 2) mesh export capabilities for seamless compatibility with existing computer graphics pipelines; 3) disentangled object representations for augmented interactivity. The core of our framework is a semantically layered 3D mesh representation that leverages panoramic images as 360° world proxies for semantic-aware world decomposition and reconstruction, enabling the generation of diverse 3D worlds. Extensive experiments demonstrate that our method achieves state-of-the-art performance in generating coherent, explorable, and interactive 3D worlds while enabling versatile applications in virtual reality, physical simulation, game development, and interactive content creation.

Sicong Liu, Yang Shu, Chenjuan Guo et al.

Learning cooperative multi-agent policy from offline multi-task data that can generalize to unseen tasks with varying numbers of agents and targets is an attractive problem in many scenarios. Although aggregating general behavior patterns among multiple tasks as skills to improve policy transfer is a promising approach, two primary challenges hinder the further advancement of skill learning in offline multi-task MARL. Firstly, extracting general cooperative behaviors from various action sequences as common skills lacks bringing cooperative temporal knowledge into them. Secondly, existing works only involve common skills and can not adaptively choose independent knowledge as task-specific skills in each task for fine-grained action execution. To tackle these challenges, we propose Hierarchical and Separate Skill Discovery (HiSSD), a novel approach for generalizable offline multi-task MARL through skill learning. HiSSD leverages a hierarchical framework that jointly learns common and task-specific skills. The common skills learn cooperative temporal knowledge and enable in-sample exploitation for offline multi-task MARL. The task-specific skills represent the priors of each task and achieve a task-guided fine-grained action execution. To verify the advancement of our method, we conduct experiments on multi-agent MuJoCo and SMAC benchmarks. After training the policy using HiSSD on offline multi-task data, the empirical results show that HiSSD assigns effective cooperative behaviors and obtains superior performance in unseen tasks.

Yuzhan Wang, Sicong Liu, Bin Guo et al.

Deep learning is reshaping mobile applications, with a growing trend of deploying deep neural networks (DNNs) directly to mobile and embedded devices to address real-time performance and privacy. To accommodate local resource limitations, techniques like weight compression, convolution decomposition, and specialized layer architectures have been developed. However, the \textit{dynamic} and \textit{diverse} deployment contexts of mobile devices pose significant challenges. Adapting deep models to meet varied device-specific requirements for latency, accuracy, memory, and energy is labor-intensive. Additionally, changing processor states, fluctuating memory availability, and competing processes frequently necessitate model re-compression to preserve user experience. To address these issues, we introduce AdaScale, an elastic inference framework that automates the adaptation of deep models to dynamic contexts. AdaScale leverages a self-evolutionary model to streamline network creation, employs diverse compression operator combinations to reduce the search space and improve outcomes, and integrates a resource availability awareness block and performance profilers to establish an automated adaptation loop. Our experiments demonstrate that AdaScale significantly enhances accuracy by 5.09%, reduces training overhead by 66.89%, speeds up inference latency by 1.51 to 6.2 times, and lowers energy costs by 4.69 times.

Haoyu Bian, Bin Guo, Sicong Liu et al.

Ubiquitous on-device heart rate sensing is vital for high-stress individuals and chronic patients. Non-contact sensing, compared to contact-based tools, allows for natural user monitoring, potentially enabling more accurate and holistic data collection. However, in open and uncontrolled mobile environments, user movement and lighting introduce. Existing methods, such as curve-based or short-range deep learning recognition based on adjacent frames, strike the optimal balance between real-time performance and accuracy, especially under limited device resources. In this paper, we present UbiHR, a ubiquitous device-based heart rate sensing system. Key to UbiHR is a real-time long-range spatio-temporal model enabling noise-independent heart rate recognition and display on commodity mobile devices, along with a set of mechanisms for prompt and energy-efficient sampling and preprocessing. Diverse experiments and user studies involving four devices, four tasks, and 80 participants demonstrate UbiHR's superior performance, enhancing accuracy by up to 74.2\% and reducing latency by 51.2\%.

Biwen Lei, Yang Li, Xinhai Liu et al.

The creation of high-quality 3D assets, a cornerstone of modern game development, has long been characterized by labor-intensive and specialized workflows. This paper presents Hunyuan3D Studio, an end-to-end AI-powered content creation platform designed to revolutionize the game production pipeline by automating and streamlining the generation of game-ready 3D assets. At its core, Hunyuan3D Studio integrates a suite of advanced neural modules (such as Part-level 3D Generation, Polygon Generation, Semantic UV, etc.) into a cohesive and user-friendly system. This unified framework allows for the rapid transformation of a single concept image or textual description into a fully-realized, production-quality 3D model complete with optimized geometry and high-fidelity PBR textures. We demonstrate that assets generated by Hunyuan3D Studio are not only visually compelling but also adhere to the stringent technical requirements of contemporary game engines, significantly reducing iteration time and lowering the barrier to entry for 3D content creation. By providing a seamless bridge from creative intent to technical asset, Hunyuan3D Studio represents a significant leap forward for AI-assisted workflows in game development and interactive media.

Fenmin Wu, Sicong Liu, Kehao Zhu et al.

The rise of mobile devices equipped with numerous sensors, such as LiDAR and cameras, has spurred the adoption of multi-modal deep intelligence for distributed sensing tasks, such as smart cabins and driving assistance. However, the arrival times of mobile sensory data vary due to modality size and network dynamics, which can lead to delays (if waiting for slower data) or accuracy decline (if inference proceeds without waiting). Moreover, the diversity and dynamic nature of mobile systems exacerbate this challenge. In response, we present a shift to \textit{opportunistic} inference for asynchronous distributed multi-modal data, enabling inference as soon as partial data arrives. While existing methods focus on optimizing modality consistency and complementarity, known as modal affinity, they lack a \textit{computational} approach to control this affinity in open-world mobile environments. AdaFlow pioneers the formulation of structured cross-modality affinity in mobile contexts using a hierarchical analysis-based normalized matrix. This approach accommodates the diversity and dynamics of modalities, generalizing across different types and numbers of inputs. Employing an affinity attention-based conditional GAN (ACGAN), AdaFlow facilitates flexible data imputation, adapting to various modalities and downstream tasks without retraining. Experiments show that AdaFlow significantly reduces inference latency by up to 79.9\% and enhances accuracy by up to 61.9\%, outperforming status quo approaches.

Sicong Liu, Wentao Zhou, Zimu Zhou et al.

There is a growing demand to deploy computation-intensive deep learning (DL) models on resource-constrained mobile devices for real-time intelligent applications. Equipped with a variety of processing units such as CPUs, GPUs, and NPUs, the mobile devices hold potential to accelerate DL inference via parallel execution across heterogeneous processors. Various efficient parallel methods have been explored to optimize computation distribution, achieve load balance, and minimize communication cost across processors. Yet their practical effectiveness in the dynamic and diverse real-world mobile environment is less explored. This paper presents a holistic empirical study to assess the capabilities and challenges associated with parallel DL inference on heterogeneous mobile processors. Through carefully designed experiments covering various DL models, mobile software/hardware environments, workload patterns, and resource availability, we identify limitations of existing techniques and highlight opportunities for cross-level optimization.

Ke Ma, Jiaqi Tang, Bin Guo et al.

Despite the growing integration of deep models into mobile terminals, the accuracy of these models declines significantly due to various deployment interferences. Test-time adaptation (TTA) has emerged to improve the performance of deep models by adapting them to unlabeled target data online. Yet, the significant memory cost, particularly in resource-constrained terminals, impedes the effective deployment of most backward-propagation-based TTA methods. To tackle memory constraints, we introduce SURGEON, a method that substantially reduces memory cost while preserving comparable accuracy improvements during fully test-time adaptation (FTTA) without relying on specific network architectures or modifications to the original training procedure. Specifically, we propose a novel dynamic activation sparsity strategy that directly prunes activations at layer-specific dynamic ratios during adaptation, allowing for flexible control of learning ability and memory cost in a data-sensitive manner. Among this, two metrics, Gradient Importance and Layer Activation Memory, are considered to determine the layer-wise pruning ratios, reflecting accuracy contribution and memory efficiency, respectively. Experimentally, our method surpasses the baselines by not only reducing memory usage but also achieving superior accuracy, delivering SOTA performance across diverse datasets, architectures, and tasks.

Sicong Liu, Bin Guo, Shiyan Luo et al.

There are many deep learning (DL) powered mobile and wearable applications today continuously and unobtrusively sensing the ambient surroundings to enhance all aspects of human lives.To enable robust and private mobile sensing, DL models are often deployed locally on resource-constrained mobile devices using techniques such as model compression or offloading.However, existing methods, either front-end algorithm level (i.e. DL model compression/partitioning) or back-end scheduling level (i.e. operator/resource scheduling), cannot be locally online because they require offline retraining to ensure accuracy or rely on manually pre-defined strategies, struggle with dynamic adaptability.The primary challenge lies in feeding back runtime performance from the back-end level to the front-end level optimization decision. Moreover, the adaptive mobile DL model porting middleware with cross-level co-adaptation is less explored, particularly in mobile environments with diversity and dynamics. In response, we introduce CrowdHMTware, a dynamic context-adaptive DL model deployment middleware for heterogeneous mobile devices. It establishes an automated adaptation loop between cross-level functional components, i.e. elastic inference, scalable offloading, and model-adaptive engine, enhancing scalability and adaptability. Experiments with four typical tasks across 15 platforms and a real-world case study demonstrate that CrowdHMTware can effectively scale DL model, offloading, and engine actions across diverse platforms and tasks. It hides run-time system issues from developers, reducing the required developer expertise.

Sicong Liu, Weiye Wu, Xiangrui Xu et al.

Foundation models have reshaped AI by unifying fragmented architectures into scalable backbones with multimodal reasoning and contextual adaptation. In parallel, the long-standing notion of AI agents, defined by the sensing-decision-action loop, is entering a new paradigm: with FMs as their cognitive core, agents transcend rule-based behaviors to achieve autonomy, generalization, and self-reflection. This dual shift is reinforced by real-world demands such as autonomous driving, robotics, virtual assistants, and GUI agents, as well as ecosystem advances in embedded hardware, edge computing, mobile deployment platforms, and communication protocols that together enable large-scale deployment. Yet this convergence collides with reality: while applications demand long-term adaptability and real-time interaction, mobile and edge deployments remain constrained by memory, energy, bandwidth, and latency. This creates a fundamental tension between the growing complexity of FMs and the limited resources of deployment environments. This survey provides the first systematic characterization of adaptive, resource-efficient agentic AI systems. We summarize enabling techniques into elastic inference, test-time adaptation, dynamic multimodal integration, and agentic AI applications, and identify open challenges in balancing accuracy-latency-communication trade-offs and sustaining robustness under distribution shifts. We further highlight future opportunities in algorithm-system co-design, cognitive adaptation, and collaborative edge deployment. By mapping FM structures, cognition, and hardware resources, this work establishes a unified perspective toward scalable, adaptive, and resource-efficient agentic AI. We believe this survey can help readers to understand the connections between enabling technologies while promoting further discussions on the fusion of agentic intelligence and intelligent agents.

Team Hunyuan3D, Bowen Zhang, Chunchao Guo et al.

Recent advances in 3D-native generative models have accelerated asset creation for games, film, and design. However, most methods still rely primarily on image or text conditioning and lack fine-grained, cross-modal controls, which limits controllability and practical adoption. To address this gap, we present Hunyuan3D-Omni, a unified framework for fine-grained, controllable 3D asset generation built on Hunyuan3D 2.1. In addition to images, Hunyuan3D-Omni accepts point clouds, voxels, bounding boxes, and skeletal pose priors as conditioning signals, enabling precise control over geometry, topology, and pose. Instead of separate heads for each modality, our model unifies all signals in a single cross-modal architecture. We train with a progressive, difficulty-aware sampling strategy that selects one control modality per example and biases sampling toward harder signals (e.g., skeletal pose) while downweighting easier ones (e.g., point clouds), encouraging robust multi-modal fusion and graceful handling of missing inputs. Experiments show that these additional controls improve generation accuracy, enable geometry-aware transformations, and increase robustness for production workflows.

Wang-Tao Zhou, Zhao Kang, Sicong Liu et al.

Event prediction tasks often handle spatio-temporal data distributed in a large spatial area. Different regions in the area exhibit different characteristics while having latent correlations. This spatial heterogeneity and correlations greatly affect the spatio-temporal distributions of event occurrences, which has not been addressed by state-of-the-art models. Learning spatial dependencies of events in a continuous space is challenging due to its fine granularity and a lack of prior knowledge. In this work, we propose a novel Graph Spatio-Temporal Point Process (GSTPP) model for fine-grained event prediction. It adopts an encoder-decoder architecture that jointly models the state dynamics of spatially localized regions using neural Ordinary Differential Equations (ODEs). The state evolution is built on the foundation of a novel Self-Adaptive Anchor Graph (SAAG) that captures spatial dependencies. By adaptively localizing the anchor nodes in the space and jointly constructing the correlation edges between them, the SAAG enhances the model's ability of learning complex spatial event patterns. The proposed GSTPP model greatly improves the accuracy of fine-grained event prediction. Extensive experimental results show that our method greatly improves the prediction accuracy over existing spatio-temporal event prediction approaches.

Diancheng Li, Dongliang Fan, Renjie Zhu et al.

Soft actuators have shown great advantages in compliance and morphology matched for manipulation of delicate objects and inspection in a confined space. There is an unmet need for a soft actuator that can provide torsional motion to e.g. enlarge working space and increase degrees of freedom. Towards this goal, we present origami-inspired soft pneumatic actuators (OSPAs) made from silicone. The prototype can output a rotation of more than one revolution (up to 435°), more significant than its counterparts. Its rotation ratio (=rotation angle/ aspect ratio) is more than 136°, about twice the largest one in other literature. We describe the design and fabrication method, build the analytical model and simulation model, and analyze and optimize the parameters. Finally, we demonstrate the potentially extensive utility of the OSPAs through their integration into a gripper capable of simultaneously grasping and lifting fragile or flat objects, a versatile robot arm capable of picking and placing items at the right angle with the twisting actuators, and a soft snake robot capable of changing attitude and directions by torsion of the twisting actuators.

Wenpei Zhu, Chenghua Lu, Qule Zheng et al.

Soft grippers are receiving growing attention due to their compliance-based interactive safety and dexterity. Hybrid gripper (soft actuators enhanced by rigid constraints) is a new trend in soft gripper design. With right structural components actuated by soft actuators, they could achieve excellent grasping adaptability and payload, while also being easy to model and control with conventional kinematics. However, existing works were mostly focused on achieving superior payload and perception with simple planar workspaces, resulting in far less dexterity compared with conventional grippers. In this work, we took inspiration from the human Metacarpophalangeal (MCP) joint and proposed a new hybrid gripper design with 8 independent muscles. It was shown that adding the MCP complexity was critical in enabling a range of novel features in the hybrid gripper, including in-hand manipulation, lateral passive compliance, as well as new control modes. A prototype gripper was fabricated and tested on our proprietary dual-arm robot platform with vision guided grasping. With very lightweight pneumatic bellows soft actuators, the gripper could grasp objects over 25 times its own weight with lateral compliance. Using the dual-arm platform, highly anthropomorphic dexterous manipulations were demonstrated using two hybrid grippers, from Tug-of-war on a rigid rod, to passing a soft towel between two grippers using in-hand manipulation. Matching with the novel features and performance specifications of the proposed hybrid gripper, the underlying modeling, actuation, control, and experimental validation details were also presented, offering a promising approach to achieving enhanced dexterity, strength, and compliance in robotic grippers.

Jiahui Cheng, Bin Guo, Jiaqi Liu et al.

Surface defect detection plays an increasingly important role in manufacturing industry to guarantee the product quality. Many deep learning methods have been widely used in surface defect detection tasks, and have been proven to perform well in defects classification and location. However, deep learning-based detection methods often require plenty of data for training, which fail to apply to the real industrial scenarios since the distribution of defect categories is often imbalanced. In other words, common defect classes have many samples but rare defect classes have extremely few samples, and it is difficult for these methods to well detect rare defect classes. To solve the imbalanced distribution problem, in this paper we propose TL-SDD: a novel Transfer Learning-based method for Surface Defect Detection. First, we adopt a two-phase training scheme to transfer the knowledge from common defect classes to rare defect classes. Second, we propose a novel Metric-based Surface Defect Detection (M-SDD) model. We design three modules for this model: (1) feature extraction module: containing feature fusion which combines high-level semantic information with low-level structural information. (2) feature reweighting module: transforming examples to a reweighting vector that indicates the importance of features. (3) distance metric module: learning a metric space in which defects are classified by computing distances to representations of each category. Finally, we validate the performance of our proposed method on a real dataset including surface defects of aluminum profiles. Compared to the baseline methods, the performance of our proposed method has improved by up to 11.98% for rare defect classes.

Mengyuan Li, Bin Guo, Jiangshan Zhang et al.

Automated Guided Vehicles (AGVs) have been widely used for material handling in flexible shop floors. Each product requires various raw materials to complete the assembly in production process. AGVs are used to realize the automatic handling of raw materials in different locations. Efficient AGVs task allocation strategy can reduce transportation costs and improve distribution efficiency. However, the traditional centralized approaches make high demands on the control center's computing power and real-time capability. In this paper, we present decentralized solutions to achieve flexible and self-organized AGVs task allocation. In particular, we propose two improved multi-agent reinforcement learning algorithms, MADDPG-IPF (Information Potential Field) and BiCNet-IPF, to realize the coordination among AGVs adapting to different scenarios. To address the reward-sparsity issue, we propose a reward shaping strategy based on information potential field, which provides stepwise rewards and implicitly guides the AGVs to different material targets. We conduct experiments under different settings (3 AGVs and 6 AGVs), and the experiment results indicate that, compared with baseline methods, our work obtains up to 47\% task response improvement and 22\% training iterations reduction.

Jianwen Luo, Sicong Liu, Chengyu Lin et al.

Supernumerary Robotics Device (SRD) is an ideal solution to provide robotic assistance in overhead manual manipulation. Since two arms are occupied for the overhead task, it is desired to have additional arms to assist us in achieving other subtasks such as supporting the far end of a long plate and pushing it upward to fit in the ceiling. In this study, a method that maps human muscle force to SRD for overhead task assistance is proposed. Our methodology is to utilize redundant DoFs such as the idle muscles in the leg to control the supporting force of the SRD. A sEMG device is worn on the operator's shank where muscle signals are measured, parsed, and transmitted to SRD for control. In the control aspect, we adopted stiffness control in the task space based on torque control at the joint level. We are motivated by the fact that humans can achieve daily manipulation merely through simple inherent compliance property in joint driven by muscles. We explore to estimate the force of some particular muscles in humans and control the SRD to imitate the behaviors of muscle and output supporting forces to accomplish the subtasks such as overhead supporting. The sEMG signals detected from human muscles are extracted, filtered, rectified, and parsed to estimate the muscle force. We use this force information as the intent of the operator for proper overhead supporting force. As one of the well-known compliance control methods, stiffness control is easy to achieve using a few of straightforward parameters such as stiffness and equilibrium point. Through tuning the stiffness and equilibrium point, the supporting force of SRD in task space can be easily controlled. The muscle force estimated by sEMG is mapped to the desired force in the task space of the SRD. The desired force is transferred into stiffness or equilibrium point to output the corresponding supporting force.

Sicong Liu, Bin Guo, Ke Ma et al.

There are many deep learning (e.g., DNN) powered mobile and wearable applications today continuously and unobtrusively sensing the ambient surroundings to enhance all aspects of human lives. To enable robust and private mobile sensing, DNN tends to be deployed locally on the resource-constrained mobile devices via model compression. The current practice either hand-crafted DNN compression techniques, i.e., for optimizing DNN-relative performance (e.g., parameter size), or on-demand DNN compression methods, i.e., for optimizing hardware-dependent metrics (e.g., latency), cannot be locally online because they require offline retraining to ensure accuracy. Also, none of them have correlated their efforts with runtime adaptive compression to consider the dynamic nature of the deployment context of mobile applications. To address those challenges, we present AdaSpring, a context-adaptive and self-evolutionary DNN compression framework. It enables the runtime adaptive DNN compression locally online. Specifically, it presents the ensemble training of a retraining-free and self-evolutionary network to integrate multiple alternative DNN compression configurations (i.e., compressed architectures and weights). It then introduces the runtime search strategy to quickly search for the most suitable compression configurations and evolve the corresponding weights. With evaluation on five tasks across three platforms and a real-world case study, experiment outcomes show that AdaSpring obtains up to 3.1x latency reduction, 4.2 x energy efficiency improvement in DNNs, compared to hand-crafted compression techniques, while only incurring <= 6.2ms runtime-evolution latency.

Sicong Liu, Junzhao Du, Kaiming Nan et al.

Recent breakthroughs in Deep Neural Networks (DNNs) have fueled a tremendously growing demand for bringing DNN-powered intelligence into mobile platforms. While the potential of deploying DNNs on resource-constrained platforms has been demonstrated by DNN compression techniques, the current practice suffers from two limitations: 1) merely stand-alone compression schemes are investigated even though each compression technique only suit for certain types of DNN layers; and 2) mostly compression techniques are optimized for DNNs' inference accuracy, without explicitly considering other application-driven system performance (e.g., latency and energy cost) and the varying resource availability across platforms (e.g., storage and processing capability). To this end, we propose AdaDeep, a usage-driven, automated DNN compression framework for systematically exploring the desired trade-off between performance and resource constraints, from a holistic system level. Specifically, in a layer-wise manner, AdaDeep automatically selects the most suitable combination of compression techniques and the corresponding compression hyperparameters for a given DNN. Thorough evaluations on six datasets and across twelve devices demonstrate that AdaDeep can achieve up to $18.6\times$ latency reduction, $9.8\times$ energy-efficiency improvement, and $37.3\times$ storage reduction in DNNs while incurring negligible accuracy loss. Furthermore, AdaDeep also uncovers multiple novel combinations of compression techniques.

Sicong Liu, Junzhao Du, Anshumali Shrivastava et al.

The remarkable success of machine learning has fostered a growing number of cloud-based intelligent services for mobile users. Such a service requires a user to send data, e.g. image, voice and video, to the provider, which presents a serious challenge to user privacy. To address this, prior works either obfuscate the data, e.g. add noise and remove identity information, or send representations extracted from the data, e.g. anonymized features. They struggle to balance between the service utility and data privacy because obfuscated data reduces utility and extracted representation may still reveal sensitive information. This work departs from prior works in methodology: we leverage adversarial learning to a better balance between privacy and utility. We design a \textit{representation encoder} that generates the feature representations to optimize against the privacy disclosure risk of sensitive information (a measure of privacy) by the \textit{privacy adversaries}, and concurrently optimize with the task inference accuracy (a measure of utility) by the \textit{utility discriminator}. The result is the privacy adversarial network (\systemname), a novel deep model with the new training algorithm, that can automatically learn representations from the raw data. Intuitively, PAN adversarially forces the extracted representations to only convey the information required by the target task. Surprisingly, this constitutes an implicit regularization that actually improves task accuracy. As a result, PAN achieves better utility and better privacy at the same time! We report extensive experiments on six popular datasets and demonstrate the superiority of \systemname compared with alternative methods reported in prior work.

Zhecheng An, Sicong Liu

Paraphrase generation is an important and challenging natural language processing (NLP) task. In this work, we propose a deep generative model to generate paraphrase with diversity. Our model is based on an encoder-decoder architecture. An additional transcoder is used to convert a sentence into its paraphrasing latent code. The transcoder takes an explicit pattern embedding variable as condition, so diverse paraphrase can be generated by sampling on the pattern embedding variable. We use a Wasserstein GAN to align the distributions of the real and generated paraphrase samples. We propose a multi-class extension to the Wasserstein GAN, which allows our generative model to learn from both positive and negative samples. The generated paraphrase distribution is forced to get closer to the positive real distribution, and be pushed away from the negative distribution in Wasserstein distance. We test our model in two datasets with both automatic metrics and human evaluation. Results show that our model can generate fluent and reliable paraphrase samples that outperform the state-of-art results, while also provides reasonable variability and diversity.

Sicong Liu, Anshumali Shrivastava, Junzhao Du et al.

The remarkable success of machine learning, especially deep learning, has produced a variety of cloud-based services for mobile users. Such services require an end user to send data to the service provider, which presents a serious challenge to end-user privacy. To address this concern, prior works either add noise to the data or send features extracted from the raw data. They struggle to balance between the utility and privacy because added noise reduces utility and raw data can be reconstructed from extracted features. This work represents a methodical departure from prior works: we balance between a measure of privacy and another of utility by leveraging adversarial learning to find a sweeter tradeoff. We design an encoder that optimizes against the reconstruction error (a measure of privacy), adversarially by a Decoder, and the inference accuracy (a measure of utility) by a Classifier. The result is RAN, a novel deep model with a new training algorithm that automatically extracts features for classification that are both private and useful. It turns out that adversarially forcing the extracted features to only conveys the intended information required by classification leads to an implicit regularization leading to better classification accuracy than the original model which completely ignores privacy. Thus, we achieve better privacy with better utility, a surprising possibility in machine learning! We conducted extensive experiments on five popular datasets over four training schemes, and demonstrate the superiority of RAN compared with existing alternatives.