Tianyuan Yu

Zeyuan Wang, Da Li, Yulin Chen et al.

Online off-policy reinforcement learning (RL) is shaped by two coupled choices: the policy class and the update rule. Gaussian policies are fast and have tractable entropy, but struggle with multimodal action distributions. Generative policies are more expressive, but often require iterative sampling or lack tractable entropy estimates. On the optimisation side, SAC-style soft policy improvement and mirror descent (MD) can be viewed as minimising different KL divergences: the former moves the policy towards a value-induced Boltzmann distribution, while the latter regularises each update against the previous policy. Combining entropy regularisation with an MD constraint is therefore attractive, as it supports exploration while stabilising policy improvement; however, the resulting target can be multimodal and is poorly matched by unimodal Gaussian policies. We propose Stochastic MeanFlow Policies (SMFP), a one-step generative policy class that maps Gaussian noise to actions through a MeanFlow transformation. This stochastic reparameterisation yields a tractable entropy surrogate and allows MeanFlow policies to be trained within off-policy mirror descent under a unified objective for exploratory yet stable improvement. Across seven MuJoCo benchmarks, SMFP improves over Gaussian and generative baselines while retaining single-step inference efficiency.

Zeyuan Wang, Da Li, Yulin Chen et al.

We introduce a one-step generative policy for offline reinforcement learning that maps noise directly to actions via a residual reformulation of MeanFlow, making it compatible with Q-learning. While one-step Gaussian policies enable fast inference, they struggle to capture complex, multimodal action distributions. Existing flow-based methods improve expressivity but typically rely on distillation and two-stage training when trained with Q-learning. To overcome these limitations, we propose to reformulate MeanFlow to enable direct noise-to-action generation by integrating the velocity field and noise-to-action transformation into a single policy network-eliminating the need for separate velocity estimation. We explore several reformulation variants and identify an effective residual formulation that supports expressive and stable policy learning. Our method offers three key advantages: 1) efficient one-step noise-to-action generation, 2) expressive modelling of multimodal action distributions, and 3) efficient and stable policy learning via Q-learning in a single-stage training setup. Extensive experiments on 73 tasks across the OGBench and D4RL benchmarks demonstrate that our method achieves strong performance in both offline and offline-to-online reinforcement learning settings. Code is available at https://github.com/HiccupRL/MeanFlowQL.

Aiming Zhang, Tianyuan Yu, Liang Bai et al.

Test-time adaptation (TTA) has gained increasing popularity due to its efficacy in addressing ``distribution shift'' issue while simultaneously protecting data privacy. However, most prior methods assume that a paired source domain model and target domain sharing the same label space coexist, heavily limiting their applicability. In this paper, we investigate a more general source model capable of adaptation to multiple target domains without needing shared labels. This is achieved by using a pre-trained vision-language model (VLM), \egno, CLIP, that can recognize images through matching with class descriptions. While the zero-shot performance of VLMs is impressive, they struggle to effectively capture the distinctive attributes of a target domain. To that end, we propose a novel method -- Context-aware Language-driven TTA (COLA). The proposed method incorporates a lightweight context-aware module that consists of three key components: a task-aware adapter, a context-aware unit, and a residual connection unit for exploring task-specific knowledge, domain-specific knowledge from the VLM and prior knowledge of the VLM, respectively. It is worth noting that the context-aware module can be seamlessly integrated into a frozen VLM, ensuring both minimal effort and parameter efficiency. Additionally, we introduce a Class-Balanced Pseudo-labeling (CBPL) strategy to mitigate the adverse effects caused by class imbalance. We demonstrate the effectiveness of our method not only in TTA scenarios but also in class generalisation tasks. The source code is available at https://github.com/NUDT-Bai-Group/COLA-TTA.

Yulin Chen, Zeyuan Wang, Tianyuan Yu et al.

The well-aligned attribute of CLIP-based models enables its effective application like CLIPscore as a widely adopted image quality assessment metric. However, such a CLIP-based metric is vulnerable for its delicate multimodal alignment. In this work, we propose \textbf{FoCLIP}, a feature-space misalignment framework for fooling CLIP-based image quality metric. Based on the stochastic gradient descent technique, FoCLIP integrates three key components to construct fooling examples: feature alignment as the core module to reduce image-text modality gaps, the score distribution balance module and pixel-guard regularization, which collectively optimize multimodal output equilibrium between CLIPscore performance and image quality. Such a design can be engineered to maximize the CLIPscore predictions across diverse input prompts, despite exhibiting either visual unrecognizability or semantic incongruence with the corresponding adversarial prompts from human perceptual perspectives. Experiments on ten artistic masterpiece prompts and ImageNet subsets demonstrate that optimized images can achieve significant improvement in CLIPscore while preserving high visual fidelity. In addition, we found that grayscale conversion induces significant feature degradation in fooling images, exhibiting noticeable CLIPscore reduction while preserving statistical consistency with original images. Inspired by this phenomenon, we propose a color channel sensitivity-driven tampering detection mechanism that achieves 91% accuracy on standard benchmarks. In conclusion, this work establishes a practical pathway for feature misalignment in CLIP-based multimodal systems and the corresponding defense method.

Tianyuan Yu, Lan Wang, Beichuan Zhang et al.

The TCP/IP protocol stack uses IP addresses for two distinct roles: identifying hosts and locating their attachment points in the network topology. This dual purpose creates a fundamental tension that has led to routing and forwarding scalability challenges throughout the history of the Internet in unicast packet delivery and, more notably, in multicast delivery. This paper reviews the evolution of routing scalability solutions over the years and makes four observations. First, map-and-encap is a recurring architectural solution shared by all scalable unicast and multicast delivery methods, developed independently across different problem contexts. Second, a new solution tends to succeed when it can bring immediate local gains to early adopters without requiring coordination across administrative domains. Third, network routing and forwarding designs that depend on external factors, such as the number of distinct end sites or even application-specific deliveries, inherently preclude an upper bound on their scalability. Fourth, today's inter-domain routing protocol, BGP, lacks a topological abstraction equivalent to an egress router within a routing domain, thereby inherently preventing a map-and-encap solution for scalability. These observations offer insights into the design of future scalable routing system architectures.

Yihang Guo, Tianyuan Yu, Liang Bai et al.

Multi-task learning (MTL) aims to build general-purpose vision systems by training a single network to perform multiple tasks jointly. While promising, its potential is often hindered by "unbalanced optimization", where task interference leads to subpar performance compared to single-task models. To facilitate research in MTL, this paper presents a systematic experimental analysis to dissect the factors contributing to this persistent problem. Our investigation confirms that the performance of existing optimization methods varies inconsistently across datasets, and advanced architectures still rely on costly grid-searched loss weights. Furthermore, we show that while powerful Vision Foundation Models (VFMs) provide strong initialization, they do not inherently resolve the optimization imbalance, and merely increasing data quantity offers limited benefits. A crucial finding emerges from our analysis: a strong correlation exists between the optimization imbalance and the norm of task-specific gradients. We demonstrate that this insight is directly applicable, showing that a straightforward strategy of scaling task losses according to their gradient norms can achieve performance comparable to that of an extensive and computationally expensive grid search. Our comprehensive analysis suggests that understanding and controlling gradient dynamics is a more direct path to stable MTL than developing increasingly complex methods.

Tianyuan Yu, Sen He, Yi-Zhe Song et al.

Graph neural networks (GNNs) have been used to tackle the few-shot learning (FSL) problem and shown great potentials under the transductive setting. However under the inductive setting, existing GNN based methods are less competitive. This is because they use an instance GNN as a label propagation/classification module, which is jointly meta-learned with a feature embedding network. This design is problematic because the classifier needs to adapt quickly to new tasks while the embedding does not. To overcome this problem, in this paper we propose a novel hybrid GNN (HGNN) model consisting of two GNNs, an instance GNN and a prototype GNN. Instead of label propagation, they act as feature embedding adaptation modules for quick adaptation of the meta-learned feature embedding to new tasks. Importantly they are designed to deal with a fundamental yet often neglected challenge in FSL, that is, with only a handful of shots per class, any few-shot classifier would be sensitive to badly sampled shots which are either outliers or can cause inter-class distribution overlapping. %Our two GNNs are designed to address these two types of poorly sampled few-shots respectively and their complementarity is exploited in the hybrid GNN model. Extensive experiments show that our HGNN obtains new state-of-the-art on three FSL benchmarks.

Zhiyi Zhang, Tianyuan Yu, Xinyu Ma et al.

Recent years have witnessed the rapid deployment of smart homes; most of them are controlled by remote servers in the cloud. Such designs raise security and privacy concerns for end users. In this paper, we describe the design of Sovereign, a home IoT system framework that provides end users complete control of their home IoT systems. Sovereign lets home IoT devices and applications communicate via application-named data and secures data directly. This enables direct, secure, one-to-one and one-to-many device-to-device communication over wireless broadcast media. Sovereign utilizes semantic names to construct usable security solutions. We implement Sovereign as a publish-subscribe-based development platform together with a prototype home IoT controller. Our preliminary evaluation shows that Sovereign provides a systematic, easy-to-use solution to user-controlled, self-contained smart homes running on existing IoT hardware without imposing noticeable overhead.