Split Federated Learning (SFL) has emerged as an efficient alternative to traditional Federated Learning (FL) by reducing client-side computation through model partitioning. However, exchanging of intermediate activations and model updates introduces significant privacy risks, especially from data reconstruction attacks that recover original inputs from intermediate representations. Existing defenses using noise injection often degrade model performance. To overcome these challenges, we present PM-SFL, a scalable and privacy-preserving SFL framework that incorporates Probabilistic Mask training to add structured randomness without relying on explicit noise. This mitigates data reconstruction risks while maintaining model utility. To address data heterogeneity, PM-SFL employs personalized mask learning that tailors submodel structures to each client's local data. For system heterogeneity, we introduce a layer-wise knowledge compensation mechanism, enabling clients with varying resources to participate effectively under adaptive model splitting. Theoretical analysis confirms its privacy protection, and experiments on image and wireless sensing tasks demonstrate that PM-SFL consistently improves accuracy, communication efficiency, and robustness to privacy attacks, with particularly strong performance under data and system heterogeneity.

Large Language Models (LLMs) have transformed natural language processing by learning from massive datasets, yet this rapid progress has also drawn legal scrutiny, as the ability to unintentionally generate copyrighted content has already prompted several prominent lawsuits. In this work, we introduce SUV (Selective Unlearning for Verbatim data), a selective unlearning framework designed to prevent LLM from memorizing copyrighted content while preserving its overall utility. In detail, the proposed method constructs a dataset that captures instances of copyrighted infringement cases by the targeted LLM. With the dataset, we unlearn the content from the LLM by means of Direct Preference Optimization (DPO), which replaces the verbatim copyrighted content with plausible and coherent alternatives. Since DPO may hinder the LLM's performance in other unrelated tasks, we integrate gradient projection and Fisher information regularization to mitigate the degradation. We validate our approach using a large-scale dataset of 500 famous books (predominantly copyrighted works) and demonstrate that SUV significantly reduces verbatim memorization with negligible impact on the performance on unrelated tasks. Extensive experiments on both our dataset and public benchmarks confirm the scalability and efficacy of our approach, offering a promising solution for mitigating copyright risks in real-world LLM applications.

Retrieval-augmented generation (RAG) agents are increasingly deployed to answer questions over local knowledge bases that cannot be centralized due to knowledge-sovereignty constraints. This results in two recurring failures in production: users do not know which agent to consult, and complex questions require evidence distributed across multiple agents. To overcome these challenges, we propose RIRS, a training-free orchestration framework to enable a multi-agent system for question answering. In detail, RIRS summarizes each agent's local corpus in an embedding space, enabling a user-facing server to route queries only to the most relevant agents, reducing latency and avoiding noisy "broadcast-to-all" contexts. For complicated questions, the server can iteratively aggregate responses to derive intermediate results and refine the question to bridge the gap toward a comprehensive answer. Extensive experiments demonstrate the effectiveness of RIRS, including its ability to precisely select agents and provide accurate responses to single-hop queries, and its use of an iterative strategy to achieve accurate, multi-step resolutions for complex queries.

Reinforcement learning with human feedback (RLHF) fine-tunes a pretrained large language model (LLM) using user preference data, enabling it to generate content aligned with human preferences. However, due to privacy concerns, users may be reluctant to share sensitive preference data. To address this, we propose utilizing Federated Learning (FL) techniques, allowing large-scale preference collection from diverse real-world users without requiring them to transmit data to a central server. Our federated RLHF methods (i.e., FedBis and FedBiscuit) encode each client's preferences into binary selectors and aggregate them to capture common preferences. In particular, FedBiscuit overcomes key challenges, such as preference heterogeneity and reward hacking, through innovative solutions like grouping clients with similar preferences to reduce heterogeneity and using multiple binary selectors to enhance LLM output quality. To evaluate the performance of the proposed methods, we establish the first federated RLHF benchmark with a heterogeneous human preference dataset. Experimental results show that by integrating the LLM with aggregated client preferences, FedBis and FedBiscuit significantly enhance the professionalism and readability of the generated content.

The advent of Large Language Models (LLMs) has significantly transformed the AI landscape, enhancing machine learning and AI capabilities. Factuality issue is a critical concern for LLMs, as they may generate factually incorrect responses. In this paper, we propose GraphEval to evaluate an LLM's performance using a substantially large test dataset. Specifically, the test dataset is retrieved from a large knowledge graph with more than 10 million facts without expensive human efforts. Unlike conventional methods that evaluate LLMs based on generated responses, GraphEval streamlines the evaluation process by creating a judge model to estimate the correctness of the answers given by the LLM. Our experiments demonstrate that the judge model's factuality assessment aligns closely with the correctness of the LLM's generated outputs, while also substantially reducing evaluation costs. Besides, our findings offer valuable insights into LLM performance across different metrics and highlight the potential for future improvements in ensuring the factual integrity of LLM outputs. The code is publicly available at https://github.com/xz-liu/GraphEval.

Model-heterogeneity federated learning (FL) is a flexible setting where a client trains a model subject to its local computation capacity. Towards the scenario, partial averaging extracts the clients' models from a global model so that the aggregation of each model parameter is identified. While existing models can only generate submodels with predefined settings established during training, our approach utilizes a trainable probabilistic masking strategy named FedMAP, enabling the dynamic creation of customized model sizes aligned with the client's budget. In detail, the clients find the best model architectures based on their local datasets and computation resources, and the FL server merges these local optimal architectures into a probabilistic mask. In the end, we attain a stable probabilistic mask, with which we can generate arbitrary models for evaluation or update the counterpart of the model parameters while training with the clients' data. Our experiments validate the effectiveness of the proposed FedMAP from two aspects: (i) It can improve the state-of-the-art approaches to heterogeneous model updates, especially for those small-size models; and (ii) We can extract the submodels whose sizes never appear in training with exceptional performance.

Large Language Models (LLMs) have transformed machine learning but raised significant legal concerns due to their potential to produce text that infringes on copyrights, resulting in several high-profile lawsuits. The legal landscape is struggling to keep pace with these rapid advancements, with ongoing debates about whether generated text might plagiarize copyrighted materials. Current LLMs may infringe on copyrights or overly restrict non-copyrighted texts, leading to these challenges: (i) the need for a comprehensive evaluation benchmark to assess copyright compliance from multiple aspects; (ii) evaluating robustness against safeguard bypassing attacks; and (iii) developing effective defense targeted against the generation of copyrighted text. To tackle these challenges, we introduce a curated dataset to evaluate methods, test attack strategies, and propose lightweight, real-time defense to prevent the generation of copyrighted text, ensuring the safe and lawful use of LLMs. Our experiments demonstrate that current LLMs frequently output copyrighted text, and that jailbreaking attacks can significantly increase the volume of copyrighted output. Our proposed defense mechanism significantly reduces the volume of copyrighted text generated by LLMs by effectively refusing malicious requests. Code is publicly available at https://github.com/xz-liu/SHIELD

Iteration based collaborative learning (CL) paradigms, such as federated learning (FL) and split learning (SL), faces challenges in training neural models over the rapidly growing yet resource-constrained edge devices. Such devices have difficulty in accommodating a full-size large model for FL or affording an excessive waiting time for the mandatory synchronization step in SL. To deal with such challenge, we propose a novel CL framework which adopts an tree-aggregation structure with an adaptive partition and ensemble strategy to achieve optimal synchronization and fast convergence at scale. To find the optimal split point for heterogeneous clients, we also design a novel partitioning algorithm by minimizing the idleness during communication and achieving the optimal synchronization between clients. In addition, a parallelism paradigm is proposed to unleash the potential of optimum synchronization between the clients and server to boost the distributed training process without losing model accuracy for edge devices. Furthermore, we theoretically prove that our framework can achieve better convergence rate than state-of-the-art CL paradigms. We conduct extensive experiments and show that our framework is 4.6× in training speed as compared with the traditional methods, without compromising training accuracy.

Multi-modal sensing systems are becoming increasingly common in real-world applications like human activity recognition (HAR). To enable knowledge sharing among individuals, Federated Learning (FL) offers a solution as a distributed machine learning paradigm that retains user data locally, thereby safeguarding privacy. However, existing heterogeneous multi-modal Federated Learning (MMFL) solutions have yet to fully utilize all the potential knowledge-sharing opportunities, as they fail to capture fundamental common knowledge that is independent of both modality and client. In this paper, we propose Federated Hierarchical Knowledge Disentanglement (FedHKD), a new sensing system for heterogeneous multi-modal federated learning. FedHKD introduces a multi-stage training paradigm based on hierarchical knowledge disentanglement at both the modality and client levels. This design enhances collaboration among modality-heterogeneous clients while maintaining low storage overhead and high adaptation flexibility to new sensing modalities. Our evaluation of two public real-world multi-modal HAR datasets and a self-collected dataset demonstrates that FedHKD outperforms state-of-the-art baselines by up to 4.85% in accuracy while saving up to 2.29x in storage. Additionally, when adapting to new sensing modalities, it reduces communication overhead by up to 4.62x.

Fair machine learning seeks to mitigate model prediction bias against certain demographic subgroups such as elder and female. Recently, fair representation learning (FRL) trained by deep neural networks has demonstrated superior performance, whereby representations containing no demographic information are inferred from the data and then used as the input to classification or other downstream tasks. Despite the development of FRL methods, their vulnerability under data poisoning attack, a popular protocol to benchmark model robustness under adversarial scenarios, is under-explored. Data poisoning attacks have been developed for classical fair machine learning methods which incorporate fairness constraints into shallow-model classifiers. Nonetheless, these attacks fall short in FRL due to notably different fairness goals and model architectures. This work proposes the first data poisoning framework attacking FRL. We induce the model to output unfair representations that contain as much demographic information as possible by injecting carefully crafted poisoning samples into the training data. This attack entails a prohibitive bilevel optimization, wherefore an effective approximated solution is proposed. A theoretical analysis on the needed number of poisoning samples is derived and sheds light on defending against the attack. Experiments on benchmark fairness datasets and state-of-the-art fair representation learning models demonstrate the superiority of our attack.

Large language models (LLMs) show amazing performance on many domain-specific tasks after fine-tuning with some appropriate data. However, many domain-specific data are privately distributed across multiple owners. Thus, this dilemma raises the interest in how to perform LLM fine-tuning in federated learning (FL). However, confronted with limited computation and communication capacities, FL clients struggle to fine-tune an LLM effectively. To this end, we introduce FedBiOT, a resource-efficient LLM fine-tuning approach to FL. Specifically, our method involves the server generating a compressed LLM and aligning its performance with the full model. Subsequently, the clients fine-tune a lightweight yet important part of the compressed model, referred to as an adapter. Notice that as the server has no access to the private data owned by the clients, the data used for alignment by the server has a different distribution from the one used for fine-tuning by clients. We formulate the problem into a bi-level optimization problem to minimize the negative effect of data discrepancy and derive the updating rules for the server and clients. We conduct extensive experiments on LLaMA-2, empirically showing that the adapter has exceptional performance when reintegrated into the global LLM. The results also indicate that the proposed FedBiOT significantly reduces resource consumption compared to existing benchmarks, all while achieving comparable performance levels.

In federated learning (FL), accommodating clients' varied computational capacities poses a challenge, often limiting the participation of those with constrained resources in global model training. To address this issue, the concept of model heterogeneity through submodel extraction has emerged, offering a tailored solution that aligns the model's complexity with each client's computational capacity. In this work, we propose Federated Importance-Aware Submodel Extraction (FIARSE), a novel approach that dynamically adjusts submodels based on the importance of model parameters, thereby overcoming the limitations of previous static and dynamic submodel extraction methods. Compared to existing works, the proposed method offers a theoretical foundation for the submodel extraction and eliminates the need for additional information beyond the model parameters themselves to determine parameter importance, significantly reducing the overhead on clients. Extensive experiments are conducted on various datasets to showcase the superior performance of the proposed FIARSE.

Due to data and information privacy concerns, data owners are not willing to share the data with others, but each of them may not have sufficient data to fine-tune a satisfactory large language model (LLM) individually. Parallelly, the LLM owners may not be willing to disclose the LLMs' details, including their architectures and parameters. Therefore, this leads to the challenge of fine-tuning an LLM on a federated learning task where the clients with task-specific data cannot obtain the complete LLM. To solve the challenge, this paper introduces FedBiOT, a method that guarantees the clients' data privacy and avoids the disclosure of an LLM. Specifically, we formulate and solve a bi-level optimization problem to ensure that the emulator distilled on the public dataset by the LLM owner can help the adaptors' local fine-tuning on clients' private datasets, regardless of the distribution drift between those datasets. Different clients' adapters are synchronized in a federated learning style, and the full model composed with the final derived adapter can achieve better performance on downstream tasks. We conduct extensive experiments on LLaMA-7B training for various federated learning tasks and witness significant improvements over existing baselines.

In the setting of federated optimization, where a global model is aggregated periodically, step asynchronism occurs when participants conduct model training by efficiently utilizing their computational resources. It is well acknowledged that step asynchronism leads to objective inconsistency under non-i.i.d. data, which degrades the model's accuracy. To address this issue, we propose a new algorithm FedaGrac, which calibrates the local direction to a predictive global orientation. Taking advantage of the estimated orientation, we guarantee that the aggregated model does not excessively deviate from the global optimum while fully utilizing the local updates of faster nodes. We theoretically prove that FedaGrac holds an improved order of convergence rate than the state-of-the-art approaches and eliminates the negative effect of step asynchronism. Empirical results show that our algorithm accelerates the training and enhances the final accuracy.

Cross-lingual natural language understanding~(NLU) aims to train NLU models on a source language and apply the models to NLU tasks in target languages, and is a fundamental task for many cross-language applications. Most of the existing cross-lingual NLU models assume the existence of parallel corpora so that words and sentences in source and target languages could be aligned. However, the construction of such parallel corpora is expensive and sometimes infeasible. Motivated by this challenge, recent works propose data augmentation or adversarial training methods to reduce the reliance on external parallel corpora. In this paper, we propose an orthogonal and novel perspective to tackle this challenging cross-lingual NLU task (i.e., when parallel corpora are unavailable). We propose to conduct multi-task learning across different tasks for mutual performance improvement on both source and target languages. The proposed multi-task learning framework is complementary to existing studies and could be integrated with existing methods to further improve their performance on challenging cross-lingual NLU tasks. Towards this end, we propose a multi-task adversarial framework for cross-lingual NLU, namely Macular. The proposed Macular includes a multi-task module and a task-specific module to infer both the common knowledge across tasks and unique task characteristics. More specifically, in the multi-task module, we incorporate a task adversarial loss into training to ensure the derivation of task-shared knowledge only by the representations. In the task-specific fine-tuning module, we extract task-specific knowledge which is not captured by the multi-task module. A task-level consistency loss is added to the training loss so that consistent predictions across a target task and an auxiliary task (i.e., the task that is the most similar to the target task) are achieved. A language adversarial loss is also incorporated so that knowledge can be transferred from source languages to target ones. To validate the effectiveness of the proposed Macular, we conduct extensive experiments on four public datasets including paraphrase identification, natural language understanding, question answering matching, and query advertisement matching. The experimental results show that the proposed Macular can outperform state-of-the-art cross-lingual NLU approaches.

In federated learning, the support of partial client participation offers a flexible training strategy, but it deteriorates the model training efficiency. In this paper, we propose a framework FedAMD to improve the convergence property and maintain flexibility. The core idea is anchor sampling, which disjoints the partial participants into anchor and miner groups. Each client in the anchor group aims at the local bullseye with the gradient computation using a large batch. Guided by the bullseyes, clients in the miner group steer multiple near-optimal local updates using small batches and update the global model. With the joint efforts from both groups, FedAMD is able to accelerate the training process as well as improve the model performance. Measured by $\epsilon$-approximation and compared to the state-of-the-art first-order methods, FedAMD achieves the convergence by up to $O(1/\epsilon)$ fewer communication rounds under non-convex objectives. In specific, we achieve a linear convergence rate under PL conditions. Empirical studies on real-world datasets validate the effectiveness of FedAMD and demonstrate the superiority of our proposed algorithm: Not only does it considerably save computation and communication costs, but also the test accuracy significantly improves.

Compared with full client participation, partial client participation is a more practical scenario in federated learning, but it may amplify some challenges in federated learning, such as data heterogeneity. The lack of inactive clients' updates in partial client participation makes it more likely for the model aggregation to deviate from the aggregation based on full client participation. Training with large batches on individual clients is proposed to address data heterogeneity in general, but their effectiveness under partial client participation is not clear. Motivated by these challenges, we propose to develop a novel federated learning framework, referred to as FedAMD, for partial client participation. The core idea is anchor sampling, which separates partial participants into anchor and miner groups. Each client in the anchor group aims at the local bullseye with the gradient computation using a large batch. Guided by the bullseyes, clients in the miner group steer multiple near-optimal local updates using small batches and update the global model. By integrating the results of the two groups, FedAMD is able to accelerate the training process and improve the model performance. Measured by $\epsilon$-approximation and compared to the state-of-the-art methods, FedAMD achieves the convergence by up to $O(1/\epsilon)$ fewer communication rounds under non-convex objectives. Empirical studies on real-world datasets validate the effectiveness of FedAMD and demonstrate the superiority of the proposed algorithm: Not only does it considerably save computation and communication costs, but also the test accuracy significantly improves.

Federated learning (FL) is an effective technique to directly involve edge devices in machine learning training while preserving client privacy. However, the substantial communication overhead of FL makes training challenging when edge devices have limited network bandwidth. Existing work to optimize FL bandwidth overlooks downstream transmission and does not account for FL client sampling. In this paper we propose GlueFL, a framework that incorporates new client sampling and model compression algorithms to mitigate low download bandwidths of FL clients. GlueFL prioritizes recently used clients and bounds the number of changed positions in compression masks in each round. Across three popular FL datasets and three state-of-the-art strategies, GlueFL reduces downstream client bandwidth by 27% on average and reduces training time by 29% on average.

Applying peer-to-peer (P2P) architecture to online video games has already attracted both academic and industrial interests, since it removes the need for expensive server maintenance. However, there are two major issues preventing the use of a P2P architecture, namely how to provide an effective distributed data storage solution, and how to tackle potential cheating behaviors. Inspired by emerging blockchain techniques, we propose a novel consensus model called Proof-of-Play (PoP) to provide a decentralized data storage system that incorporates an anti-cheating mechanism for P2P games, by rewarding players that interact with the game as intended, along with consideration of security measures to address the Nothing-at-stake Problem and the Long-range Attack. To validate our design, we utilize a game-theory model to show that under certain assumptions, the integrity of the PoP system would not be undermined due to the best interests of any user. Then, as a proof-of-concept, we developed a P2P game ( Infinity Battle ) to demonstrate how a game can be integrated with PoP in practice. Finally, experiments were conducted to study PoP in comparison with Proof-of-Work (PoW) to show its advantages in various aspects.

Traditional one-bit compressed stochastic gradient descent can not be directly employed in multi-hop all-reduce, a widely adopted distributed training paradigm in network-intensive high-performance computing systems such as public clouds. According to our theoretical findings, due to the cascading compression, the training process has considerable deterioration on the convergence performance. To overcome this limitation, we implement a sign-bit compression-based learning synchronization framework, Marsit. It prevents cascading compression via an elaborate bit-wise operation for unbiased sign aggregation and its specific global compensation mechanism for mitigating compression deviation. The proposed framework retains the same theoretical convergence rate as noncompression mechanisms. Experimental results demonstrate that Marsit reduces up to 35% training time while preserving the same accuracy as training without compression.