LLM's code generation capabilities have yielded substantial improvements in the effectiveness of programming tasks. However, LLM-generated code still suffers from compilation and runtime errors. Existing offline preference optimization methods primarily focus on enhancing LLMs' coding abilities using pass/fail signals in the preference data, overlooking the deep-level error types in the failed codes. To address this, we propose Adaptively Progressive Preference Optimization (AP2O) for coding (i.e., AP2O-Coder), a method that guides LLMs adaptively and methodically to reduce code errors for code generation. Specifically, we construct an error notebook from failed codes and progressively optimize the LLM to correct errors type by type. Furthermore, we adaptively replay error types to tailor to the LLM's evolving weaknesses throughout training. Through extensive experiments on both code and general LLMs (Llama, Qwen, and DeepSeek series) with parameters ranging from 0.5B to 34B, our AP2O-Coder improves code generation performance by up to 3% in pass@k while using less preference data.

Zhezheng Hao, Hong Wang, Jian Luo, Jianqing Zhang, Yuyan Zhou, Qiang Lin, Can Wang, Hande Dong, Jiawei Chen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Mobile GUI agents have attracted tremendous research participation recently.Traditional approaches to mobile agent training rely on centralized data collection, leading to high cost and limited scalability.Distributed training utilizing federated learning offers an alternative by harnessing real-world user data, providing scalability and reducing costs.However, pivotal challenges, including the absence of standardized benchmarks, hinder progress in this field.To tackle the challenges, we introduce FedMABench, the first benchmark for federated training and evaluation of mobile GUI agents, specifically designed for heterogeneous scenarios.FedMABench features 6 datasets with 30+ subsets, 8 federated algorithms, 10+ base models, and over 800 apps across 5 categories, providing a comprehensive framework for evaluating mobile agents across diverse environments.Through extensive experiments, we uncover several key insights: federated algorithms consistently outperform local training; the distribution of specific apps plays a crucial role in heterogeneity; and, even apps from distinct categories can exhibit correlations during training.

The rise of generative APIs has fueled interest in privacy-preserving synthetic data generation. While the Private Evolution (PE) algorithm generates Differential Privacy (DP) synthetic images using diffusion model APIs, it struggles with few-shot private data due to the limitations of its DP-protected similarity voting approach. In practice, the few-shot private data challenge is particularly prevalent in specialized domains like healthcare and industry. To address this challenge, we propose a novel API-assisted algorithm, Private Contrastive Evolution (PCEvolve), which iteratively mines inherent inter-class contrastive relationships in few-shot private data beyond individual data points and seamlessly integrates them into an adapted Exponential Mechanism (EM) to optimize DP’s utility in an evolution loop. We conduct extensive experiments on four specialized datasets, demonstrating that PCEvolve outperforms PE and other API-assisted baselines. These results highlight the potential of leveraging API access with private data for quality evaluation, enabling the generation of high-quality DP synthetic images and paving the way for more accessible and effective privacy-preserving generative API applications. Our code is available at https://github.com/TsingZ0/PCEvolve.

Substantial quantity and high quality are the golden rules of making a good training dataset with sample privacy protection equally important. Generating synthetic samples that resemble high-quality private data while ensuring Differential Privacy (DP), a formal privacy guarantee, promises scalability and practicality. However, existing methods relying on pre-trained models for data synthesis often struggle in data-deficient scenarios, suffering from limited sample size, inevitable generation noise and existing pre-trained model bias. To address these challenges, we propose a novel contr**A**stive private data **S**ynthesis via **W**eighted multiple **P**re-trained generative models framework, named as **WASP**. WASP utilizes limited private samples for more accurate private data distribution estimation via a Top-*Q* voting mechanism, and leverages low-quality synthetic samples for contrastive generation via collaboration among dynamically weighted multiple pre-trained models. Extensive experiments on 6 well-developed datasets with 6 open-source and 3 closed-source PLMs demonstrate the superiority of WASP in improving model performance over diverse downstream tasks. Code is available at https://github.com/LindaLydia/WASP.

As AI evolves, collaboration among heterogeneous models helps overcome data scarcity by enabling knowledge transfer across institutions and devices.Traditional Federated Learning (FL) only supports homogeneous models, limiting collaboration among clients with heterogeneous model architectures.To address this, Heterogeneous Federated Learning (HtFL) methods are developed to enable collaboration across diverse heterogeneous models while tackling the data heterogeneity issue at the same time.However, a comprehensive benchmark for standardized evaluation and analysis of the rapidly growing HtFL methods is lacking.Firstly, the highly varied datasets, model heterogeneity scenarios, and different method implementations become hurdles to making easy and fair comparisons among HtFL methods.Secondly, the effectiveness and robustness of HtFL methods are under-explored in various scenarios, such as the medical domain and sensor signal modality.To fill this gap, we introduce the first Heterogeneous Federated Learning Library (HtFLlib), an easy-to-use and extensible framework that integrates multiple datasets and model heterogeneity scenarios, offering a robust benchmark for research and practical applications.Specifically, HtFLlib integrates (1) 12 datasets spanning various domains, modalities, and data heterogeneity scenarios; (2) 40 model architectures, ranging from small to large, across three modalities;(3) a modularized and easy-to-extend HtFL codebase with implementations of 10 representative HtFL methods; and (4) systematic evaluations in terms of accuracy, convergence, computation costs, and communication costs.We emphasize the advantages and potential of state-of-the-art HtFL methods and hope that HtFLlib will catalyze advancing HtFL research and enable its broader applications.The code is released at https://github.com/TsingZ0/HtFLlib.

Amid the ongoing advancements in Federated Learning (FL), a machine learning paradigm that allows collaborative learning with data privacy protection, personalized FL (pFL) has gained significant prominence as a research direction within the FL domain. Whereas traditional FL (tFL) focuses on jointly learning a global model, pFL aims to balance each client's global and personalized goals in FL settings. To foster the pFL research community, we started and built PFLlib, a comprehensive pFL library with an integrated benchmark platform. In PFLlib, we implemented 37 state-of-the-art FL algorithms (8 tFL algorithms and 29 pFL algorithms) and provided various evaluation environments with three statistically heterogeneous scenarios and 24 datasets. At present, PFLlib has gained more than 1600 stars and 300 forks on GitHub.

Recently, Heterogeneous Federated Learning (HtFL) has attracted attention due to its ability to support heterogeneous models and data. To reduce the high communication cost of transmitting model parameters, a major challenge in HtFL, prototype-based HtFL methods are proposed to solely share class representatives, a.k.a, prototypes, among heterogeneous clients while maintaining the privacy of clients’ models. However, these prototypes are naively aggregated into global prototypes on the server using weighted averaging, resulting in suboptimal global knowledge which negatively impacts the performance of clients. To overcome this challenge, we introduce a novel HtFL approach called FedTGP, which leverages our Adaptive-margin-enhanced Contrastive Learning (ACL) to learn Trainable Global Prototypes (TGP) on the server. By incorporating ACL, our approach enhances prototype separability while preserving semantic meaning. Extensive experiments with twelve heterogeneous models demonstrate that our FedTGP surpasses state-of-the-art methods by up to 9.08% in accuracy while maintaining the communication and privacy advantages of prototype-based HtFL. Our code is available at https://github.com/TsingZ0/FedTGP.

Heterogeneous Federated Learning (HtFL) enables collaborative learning on multiple clients with different model architectures while preserving privacy. Despite recent research progress, knowledge sharing in HtFL is still difficult due to data and model heterogeneity. To tackle this issue, we leverage the knowledge stored in public pretrained generators and propose a new upload-efficient knowledge transfer scheme called Federated Knowledge-Transfer Loop (Fed-KTL). Our FedKTL can produce client-task-related prototypical image-vector pairs via the generator's inference on the server. With these pairs, each client can transfer preexisting knowledge from the generator to its local model through an additional supervised local task. We conduct extensive experiments on four datasets under two types of data heterogeneity with 14 kinds of models including CNNs and ViTs. Results show that our upload-efficient FedKTL surpasses seven state-of-the-art methods by up to 7.31% in accuracy. Moreover, our knowledge transfer scheme is applicable in scenarios with only one edge client. Code: https://github.com/TsingZ0/FedKTL

Data-generation based zero-shot learning, although effective in training Small Task-specific Models (STMs) via synthetic datasets generated by Pre-trained Language Models (PLMs), is often limited by the low quality of such synthetic datasets.Previous solutions have primarily focused on single PLM settings, where synthetic datasets are typically restricted to specific sub-spaces and often deviate from real-world distributions, leading to severe distribution bias.To mitigate such bias, we propose FuseGen, a novel data-generation based zero-shot learning framework that introduces a new criteria for subset selection from synthetic datasets via utilizing multiple PLMs and trained STMs.The chosen subset provides in-context feedback to each PLM, enhancing dataset quality through iterative data generation.Trained STMs are then used for sample re-weighting as well, further improving data quality.Extensive experiments across diverse tasks demonstrate that FuseGen substantially outperforms existing methods, highly effective in boosting STM performance in a PLM-agnostic way. 1

A key challenge in federated learning (FL) is the statistical heterogeneity that impairs the generalization of the global model on each client. To address this, we propose a method Federated learning with Adaptive Local Aggregation (FedALA) by capturing the desired information in the global model for client models in personalized FL. The key component of FedALA is an Adaptive Local Aggregation (ALA) module, which can adaptively aggregate the downloaded global model and local model towards the local objective on each client to initialize the local model before training in each iteration. To evaluate the effectiveness of FedALA, we conduct extensive experiments with five benchmark datasets in computer vision and natural language processing domains. FedALA outperforms eleven state-of-the-art baselines by up to 3.27% in test accuracy. Furthermore, we also apply ALA module to other federated learning methods and achieve up to 24.19% improvement in test accuracy. Code is available at https://github.com/TsingZ0/FedALA.

Federated Learning (FL) is popular for its privacy-preserving and collaborative learning capabilities. Recently, personalized FL (pFL) has received attention for its ability to address statistical heterogeneity and achieve personalization in FL. However, from the perspective of feature extraction, most existing pFL methods only focus on extracting global or personalized feature information during local training, which fails to meet the collaborative learning and personalization goals of pFL. To address this, we propose a new pFL method, named GPFL, to simultaneously learn global and personalized feature information on each client. We conduct extensive experiments on six datasets in three statistically heterogeneous settings and show the superiority of GPFL over ten state-of-the-art methods regarding effectiveness, scalability, fairness, stability, and privacy. Besides, GPFL mitigates overfitting and outperforms the baselines by up to 8.99% in accuracy.

Recently, personalized federated learning (pFL) has attracted increasing attention in privacy protection, collaborative learning, and tackling statistical heterogeneity among clients, e.g., hospitals, mobile smartphones, etc. Most existing pFL methods focus on exploiting the global information and personalized information in the client-level model parameters while neglecting that data is the source of these two kinds of information. To address this, we propose the Federated Conditional Policy (FedCP) method, which generates a conditional policy for each sample to separate the global information and personalized information in its features and then processes them by a global head and a personalized head, respectively. FedCP is more fine-grained to consider personalization in a sample-specific manner than existing pFL methods. Extensive experiments in computer vision and natural language processing domains show that FedCP outperforms eleven state-of-the-art methods by up to 6.69%. Furthermore, FedCP maintains its superiority when some clients accidentally drop out, which frequently happens in mobile settings. Our code is public at https://github.com/TsingZ0/FedCP.

Recently, federated learning (FL) is popular for its privacy-preserving and collaborative learning abilities. However, under statistically heterogeneous scenarios, we observe that biased data domains on clients cause a representation bias phenomenon and further degenerate generic representations during local training, i.e., the representation degeneration phenomenon. To address these issues, we propose a general framework Domain Bias Eliminator (DBE) for FL. Our theoretical analysis reveals that DBE can promote bi-directional knowledge transfer between server and client, as it reduces the domain discrepancy between server and client in representation space. Besides, extensive experiments on four datasets show that DBE can greatly improve existing FL methods in both generalization and personalization abilities. The DBE-equipped FL method can outperform ten state-of-the-art personalized FL methods by a large margin. Our code is public at https://github.com/TsingZ0/DBE.

FroBot can locomote both on land and underwater based on its dual swing-legs propulsion mechanism. This paper presents the dynamic model, experimental studies, and control strategies of FroBot underwater. In this work, an experimental setup consisting of two-degree-of-freedom(2DOF) robotic swing-legs is built to study the model of FroBot underwater. We first improve the dynamic model of caudal fins based on the Morison equation. Combined with experimental data, we optimize the model parameters and then obtain the optimal control strategy of uniform swing. In addition, we apply the CPGs control strategy and improve it based on the FroBot model. These two control strategies have their advantages and demonstrate the potential for future use in control applications.

This paper describes a novel amphibious robot, which adopts a dual-swing-legs propulsion mechanism, proposing a new locomotion mode. The robot is called FroBot, since its structure and locomotion are similar to frogs. Our inspiration comes from the frog scooter and breaststroke. Based on its swing leg mechanism, an unusual universal wheel structure is used to generate propulsion on land, while a pair of flexible caudal fins functions like the foot flippers of a frog to generate similar propulsion underwater. On the basis of the prototype design and the dynamic model of the robot, some locomotion control simulations and experiments were conducted for the purpose of adjusting the parameters that affect the propulsion of the robot. Finally, a series of underwater experiments were performed to verify the design feasibility of FroBot and the rationality of the control algorithm.

With the ever increasing number of connected devices and the over abundance of data generated by these devices, data privacy has become a critical concern in the Internet of Things (IoT). One promising privacy-preservation approach is Attribute-Based Encryption (ABE), a public key encryption scheme that enables fine-grained access control, scalable key management and flexible data distribution. This paper presents an in-depth performance evaluation of ABE that focuses on execution time, data and network overhead, energy consumption, and CPU and memory usage. We evaluate two major types of ABE, Key-Policy Attribute-Based Encryption (KP-ABE) and Ciphertext-Policy Attribute-Based Encryption (CP-ABE), on different classes of mobile devices including a laptop and a smartphone. To the best of our knowledge, this is the first comprehensive study of ABE dedicated solely to its performance. Our results provide insights into important practical issues of ABE, including what computing resources ABE requires in heterogeneous environments, at what cost ABE offers benefits, and under what situations ABE is best suited for use in the IoT.

Automatic extraction of roads from images of complex urban areas is a very difficult task due to the occlusions and shadows of contextual objects, and complicated road structures. As light detection and ranging (LiDAR) data explicitly contain direct 3-D information of the urban scene and are less affected by occlusions and shadows, they are a good data source for road detection. This paper proposes to use multiple features to detect road centerlines from the remaining ground points after filtering. The main idea of our method is to effectively detect smooth geometric primitives of potential road centerlines and to separate the connected nonroad features (parking lots and bare grounds) from the roads. The method consists of three major steps, i.e., spatial clustering based on multiple features using an adaptive mean shift to detect the center points of roads, stick tensor voting to enhance the salient linear features, and a weighted Hough transform to extract the arc primitives of the road centerlines. In short, we denote our method as Mean shift, Tensor voting, Hough transform (MTH). We evaluated the method using the Vaihingen and Toronto data sets from the International Society for Photogrammetry and Remote Sensing Test Project on Urban Classification and 3-D Building Reconstruction. The completeness of the extracted road network on the Vaihingen data and the Toronto data are 81.7% and 72.3%, respectively, and the correctness are 88.4% and 89.2%, respectively, yielding the best performance compared with template matching and phase-coded disk methods.

We design a content-centric privacy scheme for Information-Centric Networking (ICN). We enhance ICN's ability to support data confidentiality by introducing attribute-based encryption into ICN and making it specific to the data attributes. Our approach is unusual in that it preserves ICN's goal to decouple publishers and subscribers for greater data accessibility, scalable multiparty communication and efficient data distribution. Inspired by application-layer publish-subscribe, we enable fine-grained access control with more expressive policies. Moreover, we propose an attribute-based routing scheme that offers interest confidentiality. A prototype system is implemented based on CCNx, a popular open source version of ICN, to showcase privacy preservation in Smart Neighborhood and Smart City applications.

Mind evolutionary computation (MEC) is a new approach of evolutionary computation (EC). It is proved that MEC has much higher computing efficiency and convergence ability than genetic algorithms (GAs). This is because of using operation similartaxis and dissimilation rather than crossover and mutation operators in GA. The paper analyzes the influence of type of the probability density function on similartaxis in MEC. We get theoretically the relation among similartaxis calculated amount, the parameters of probability density function of scattering individuals, the size of group, the precision of solution and the distance between initial searching position and local optimum. The experiment shows that the analysis method proposed in the paper is reasonable. The analysis and experiment also shows that using different types of probability density functions doesn't make much change on similartaxis searching performance.

Mind evolutionary computation (MEC) is a new approach to evolutionary computation (EC). This paper presents a new dissimilation strategy using rejected regions, which can avoid searching repeatedly, so that the capability of MEC to search globally in dissimilation is enhanced. Experimental results show that basic MEC has improved considerately compared with a genetic algorithm (GA), and that the MEC dissimilation strategy using rejected regions has also advanced a lot. The reason for this is that, in the modified MEC, the regions searched in similartaxis are recorded, so that, in dissimilation, the scope of scattered individuals is reduced to the whole solution space, excluding the rejected regions. Therefore, the regions explored in dissimilation have never been searched before, and the search scope is diminished accordingly, while the capability of MEC to search globally in dissimilation is enhanced and repeated searching is avoided. It is the memory mechanism of MEC that makes the dissimilation strategy of rejected regions possible, so the probability that the individuals are scattered in the region of the global optimum has greatly increased, the calculated amount and the average evaluation time are decreased, and population convergence can be implemented in fewer generations.

Mind evolutionary computation (MEC) is a new approach to evolutionary computation (EC). It is proved that MEC has much higher computational efficiency and convergence ability than genetic algorithms (GAs). This is because MEC uses the operations of similartaxis and dissimilation rather than the crossover and mutation operators used in GAs, and also because of the different performance mechanisms from GAs, the memory mechanism, the evolutionary directional mechanism and the harmonizing mechanism between exploitation and exploration. This paper presents a new dissimilation strategy using rejected regions, which can avoid searching repeatedly, so that the capability of MEC to search globally in dissimilation is enhanced. It is the memory mechanism of MEC that makes the dissimilation strategy of rejected regions possible.