Zhuo Chen, Minghao Li, Xiaoqian Ma, Siqi Fan, Xiusheng Huang, Zhang Liujie, Weihang Chen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Real-life deployment of federated Learning (FL) often faces non-IID data, which leads to poor accuracy and slow convergence. Personalized FL (pFL) tackles these issues by tailoring local models to individual data sources and using weighted aggregation methods for client-specific learning. However, existing pFL methods often fail to provide each local model with global knowledge on demand while maintaining low computational overhead. Additionally, local models tend to over-personalize their data during the training process, potentially dropping previously acquired global information. We propose FLAYER, a novel layer-wise learning method for pFL that optimizes local model personalization performance. FLAYER considers the different roles and learning abilities of neural network layers of individual local models. It incorporates global information for each local model as needed to initialize the local model cost-effectively. It then dynamically adjusts learning rates for each layer during local training, optimizing the personalized learning process for each local model while preserving global knowledge. Additionally, to enhance global representation in pFL, FLAYER selectively uploads parameters for global aggregation in a layer-wise manner. We evaluate FLAYER on four representative datasets in computer vision and natural language processing domains. Compared to eight state-of-the-art pFL methods, FLAYER improves the inference accuracy, on average, by 5.20% (up to 14.29%). Code is available at https://github.com/lancasterJie/FLAYER/.

Articulated objects are ubiquitous in daily life. In this paper, we present DexSim2Real$^{\mathbf{2}}$, a novel framework for goal-conditioned articulated object manipulation. The core of our framework is constructing an explicit world model of unseen articulated objects through active interactions, which enables sampling-based model predictive control to plan trajectories achieving different goals without requiring demonstrations or RL. It first predicts an interaction using an affordance network trained on self-supervised interaction data or videos of human manipulation. After executing the interactions on the real robot to move the object parts, we propose a novel modeling pipeline based on 3D AIGC to build a digital twin of the object in simulation from multiple frames of observations. For dexterous hands, we utilize eigengrasp to reduce the action dimension, enabling more efficient trajectory searching. Experiments validate the framework's effectiveness for precise manipulation using a suction gripper, a two-finger gripper and two dexterous hands. The generalizability of the explicit world model also enables advanced manipulation strategies like manipulating with tools.

Vision-based tactile sensors have drawn increasing interest in the robotics community. However, traditional lens-based designs impose minimum thickness constraints on these sensors, limiting their applicability in space-restricted settings. In this article, we propose ThinTact, a novel lensless vision-based tactile sensor with a sensing field of over 200 mm${}^{2}$and a thickness of less than 10 mm. ThinTact utilizes the mask-based lensless imaging technique to map the contact information to CMOS signals. To ensure real-time tactile sensing, we propose a real-time lensless reconstruction algorithm that leverages a frequency-spatial-domain joint filter based on discrete cosine transform. This algorithm achieves computation significantly faster than existing optimization-based methods. In addition, to improve the sensing quality, we develop a mask optimization method based on the generic algorithm and the corresponding system matrix calibration algorithm. We evaluate the performance of our proposed lensless reconstruction and tactile sensing through qualitative and quantitative experiments. Furthermore, we demonstrate ThinTact's practical applicability in diverse applications, including texture recognition and contact-rich object manipulation.

Visuotactile sensors can provide rich contact information, having great potential in contact-rich manipulation tasks with reinforcement learning (RL) policies. Sim2Real technique tackles the challenge of RL's reliance on a large amount of interaction data. However, most Sim2Real methods for manipulation tasks with visuotactile sensors rely on rigid-body physics simulation, which fails to simulate the real elastic deformation precisely. Moreover, these methods do not exploit the characteristic of tactile signals for designing the network architecture. In this paper, we build a general-purpose Sim2Real protocol for manipulation policy learning with marker-based visuotactile sensors. To improve the simulation fidelity, we employ an FEM-based physics simulator that can simulate the sensor deformation accurately and stably for arbitrary geometries. We further propose a novel tactile feature extraction network that directly processes the set of pixel coordinates of tactile sensor markers and a self-supervised pre-training strategy to improve the efficiency and generalizability of RL policies. We conduct extensive Sim2Real experiments on the peg-in-hole task to validate the effectiveness of our method. And we further show its generalizability on additional tasks including plug adjustment and lock opening. The protocol, including the simulator and the policy learning framework, will be open-sourced for community usage.

Accurately manipulating articulated objects is a challenging yet important task for real robot applications. In this paper, we present a novel framework called Sim2Real2to enable the robot to manipulate an unseen articulated object to the desired state precisely in the real world with no human demonstrations. We leverage recent advances in physics simulation and learning-based perception to build the interactive explicit physics model of the object and use it to plan a long-horizon manipulation trajectory to accomplish the task. However, the interactive model cannot be correctly estimated from a static observation. Therefore, we learn to predict the object affordance from a single-frame point cloud, control the robot to actively interact with the object with a one-step action, and capture another point cloud. Further, the physics model is constructed from the two point clouds. Experimental results show that our framework achieves about 70% manipulations with < 30% relative error for common articulated objects, and 30% manipulations for difficult objects. Our proposed framework also enables advanced manipulation strategies, such as manipulating with different tools. Code and videos are available on our project webpage: https://ttimelord.github.io/Sim2Real2-site/

Transparent objects are common in daily life. However, depth sensing for transparent objects remains a challenging problem. While learning-based methods can leverage shape priors to improve the sensing quality, the labor-intensive data collection in real world and the sim-to-real domain gap restrict these methods' scalability. In this paper, we propose a method to finetune a stereo network with sparse depth labels automatically collected using a probing system with tactile feedback. We present a novel utility function to evaluate the benefit of touches. By approximating and optimizing the utility function, we can optimize the probing locations given a fixed touching budget to better improve the network's performance on real objects. We further combine tactile depth supervision with a confidence-based regularization to prevent over-fitting during finetuning. To evaluate the effectiveness of our method, we construct a real-world dataset including both diffuse and transparent objects. Experimental results on this dataset show that our method can significantly improve real-world depth sensing accuracy, especially for transparent objects.

Sealing mechanisms are at the heart of any drilling, completion, or production system and are the primary components on which the functional success and longevity of the system rest. Element extrusion and cracking are the earliest failure modes to indicate leakage on the elastomers. With the trend towards higher operating temperatures and pressures of the well completion systems, more accurate and reliable methods are needed to identify the failure modes of the elastomer. Currently, elastomer failure type tests are carried out by a trained human supervisor. This method is time consuming and inefficient in inspecting a large quantity of elastomers. Furthermore, this manual approach depends on the human examiner's knowledge and experience, which is subjective and may be prone to errors. To overcome these issues, we employ Mask R-CNN with a modified loss function to identify the failure types and localize failure regions of elastomers of the well completion systems. To train Mask R-CNN for crack and extrusion detection, a total of 70 elastomer images with different types of failure conditions are collected from real-world environments by the oil and gas company. The VGG annotator is used to annotate the labels of the elastomer failure types and positions with polygonal regions in the images. Diverse image augmentation techniques are also used to further enhance performance. The mean average precision results are used to indicate model performance with different datasets. The empirical results indicate the robustness and efficiency of Mask R-CNN in performing failure detection and segmentation of elastomers.

Audit log contains the trace of different activities in computing systems, which makes it critical for security management, censorship, and forensics. However, experienced attackers may delete or modify the audit log after their attacks, which makes the audit log unavailable in attack investigation. In this article, we focus on the log integrity audit in the same domain, in which a number of servers update audit logs for a single or several organizations as an alliance. We propose a distributed efficient log integrity audit framework, called DELIA, which employs the distributed ledger technique to protect audit information, and utilizes the idea of state channel to improve the throughput of distributed ledger. To generate stable state from the rapidly-updated logs in the domain, we propose a log state generation scheme, which not only generates state suitable for audit logs, but also enables mutual supervision within the domain. To overcome the high latency in existing state channel schemes, we propose a hierarchal multi-party state channel scheme, which makes the latency in our framework independent of the number of servers in the domain. We implement DELIA on Ethereum and evaluate its performance. The results show that our framework is efficient and secure in practice.

Clients (e.g., browsers) and servers require public key certificates to establish secure connections. When a client accesses a server, it needs to check the signature, expiration time, and revocation status of the certificate to determine whether the server is reliable. The existing solutions for checking certificate status either have a long update cycle (e.g., CRL, CRLite) or violate clients' privacy (e.g., OCSP, CCSP), and these solutions also have the problem of trust concentration. In this paper, we present PROCESS, an online privacy-preserving on-chain certificate status service based on the blockchain architecture, which can ensure decentralized trust and provide privacy protection for clients. Specifically, we design Counting Garbled Bloom Filter (CGBF) that supports efficient queries and BlockOriented Revocation List (BORL) to update CGBF timely in the blockchain. With CGBF, we design a privacy-preserving protocol to protect clients' privacy when they check the certificate statuses from the blockchain nodes. Finally, we conduct experiments and compare PROCESS with another blockchain-based solution to demonstrate that PROCESS is suitable in practice.

Scholars' homepages are important places to show personal research interest and academic achievement through the Web. However, according to our observation, only a small portion of scholars update their publications and related events on their homepages in time. In this paper, we propose a homepage augmentation technique, which automatically shows the newest academic events related to a scholar on his/her homepage. Specifically, we model the relations between homepages and the events collected from the Web as a complex heterogenous network, and propose an Embedding-based Heterogenous random Walk algorithm, namely EHWalk, to predict the links between homepages and events. Compared with existing embedding-based link prediction algorithms, EHWalk supports more efficient modeling of complex heterogenous relations in a dynamically changing network, which helps link the massive new updated events to homepages precisely and efficiently. Comprehensive experiments on a real-world dataset are conducted and the results show that our algorithm can achieve both good effectiveness and efficiency for real-world deployment.

Most of the proposed person re-identification algorithms conduct supervised training and testing on single labeled datasets with small size, so directly deploying these trained models to a large-scale real-world camera network may lead to poor performance due to underfitting. It is challenging to incrementally optimize the models by using the abundant unlabeled data collected from the target domain. To address this challenge, we propose an unsupervised incremental learning algorithm, TFusion, which is aided by the transfer learning of the pedestrians' spatio-temporal patterns in the target domain. Specifically, the algorithm firstly transfers the visual classifier trained from small labeled source dataset to the unlabeled target dataset so as to learn the pedestrians' spatial-temporal patterns. Secondly, a Bayesian fusion model is proposed to combine the learned spatio-temporal patterns with visual features to achieve a significantly improved classifier. Finally, we propose a learning-to-rank based mutual promotion procedure to incrementally optimize the classifiers based on the unlabeled data in the target domain. Comprehensive experiments based on multiple real surveillance datasets are conducted, and the results show that our algorithm gains significant improvement compared with the state-of-art cross-dataset unsupervised person re-identification algorithms.