Kangwen Zhao, Jianfeng Cai, Jinhua Zhu, Ruopei Sun, Dongyun Xue, Wengang Zhou, Li Li, Houqiang Li. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

Federated learning (FL) is a distributed machine learning technique that guarantees the privacy of user data. However, FL has been shown to be vulnerable to gradient leakage attacks (GLA), which have the ability to reconstruct private training data from public gradients with high probability. These attacks are either analytic-based, requiring modification of the FL model, or optimization-based, requiring long convergence times and failing to effectively address the challenge of dealing with highly compressed gradients in practical FL systems. This paper presents a pioneering generation-based GLA method called FGLA that can reconstruct batches of user data without the need for the optimization process. We specifically design a feature separation technique that first extracts the features of each sample in a batch and then directly generates the user data. Our extensive experiments on multiple image datasets show that FGLA can reconstruct user images in seconds with a batch size of 256 from highly compressed gradients (0.8% compression ratio or higher), thereby significantly outperforming state-of-the-art methods.

Federated learning (FL) is a distributed machine learning technology that preserves data privacy. However, it has been shown to be vulnerable to gradient leakage attacks (GLA), which can reconstruct private training data from public gradients with an overwhelming probability. Nevertheless, these attacks either require modification of the FL model (analytics-based) or take a long time to converge (optimization-based) and fail in dealing with highly compressed gradients in practical FL systems. In this paper, we pioneer a generation-based GLA method called FGLA that can reconstruct batches of user data, forgoing the optimization process. Specifically, we design a feature separation technique that extracts the feature of each data in a batch and then generates user data directly. Extensive experiments on multiple image datasets demonstrate that FGLA can reconstruct user images in milliseconds with a batch size of 256 from highly compressed gradients (0.8% compression ratio or higher), thus substantially outperforming state-of-the-art methods.

As a popular game around the world, Snakes has multiple modes with different settings. In this work, we are dedicated to the 3v3 Snakes, which is characterized by a complex mixture of competition and cooperation. To address this mode of Snakes, most existing AI agents adopt rule based methods, which achieve limited performance due to human’s oversight of some special circumstances. Inspired by the superiority of multi-agent reinforcement learning (MARL), we propose a rule-enhanced multi-agent reinforcement learning algorithm and build a 3v3 Snakes AI. Specifically, we introduce the territory matrix which is commonly utilized in rule based methods to the state features and mask the illegal actions through designed rules. The relationships of individual-team and friends-foes are also merged into reward design. Trained with Distributed PPO and self-play on a single GeForce RTX 2080 GPU for twenty-four hours, our AI achieves state-of-the-art performance and beats human players. On JIDI platform, our agent outperforms the other 132 participating agents and ranks the first for more than 20 consecutive days.