Large language model (LLM) agents have demonstrated strong capabilities across diverse domains, yet automated agent design remains a significant challenge. Current automated agent design approaches are often constrained by limited search spaces that primarily optimize workflows but fail to integrate crucial human-designed components like memory, planning, and tool use. Furthermore, these methods are hampered by high evaluation costs, as evaluating even a single new agent on a benchmark can require tens of dollars. The difficulty of this exploration is further exacerbated by inefficient search strategies that struggle to navigate the large design space effectively, making the discovery of novel agents a slow and resource-intensive process. To address these challenges, we propose AgentSwift, a novel framework for automated agent design. We formalize a hierarchical search space that jointly models agentic workflow and composable functional components. This structure moves beyond optimizing workflows alone by co-optimizing functional components, which enables the discovery of more complex and effective agent architectures. To make exploration within this expansive space feasible, we mitigate high evaluation costs by training a value model on a high-quality dataset, generated via a novel strategy combining combinatorial coverage and balanced Bayesian sampling for low-cost evaluation. Guiding the entire process is a hierarchical Monte Carlo Tree Search (MCTS) strategy, which is informed by uncertainty to efficiently navigate the search space. Evaluated across a comprehensive set of seven benchmarks spanning embodied, math, web, tool, and game domains, AgentSwift discovers agents that achieve an average performance gain of 8.34\% over both existing automated agent search methods and manually designed agents. Moreover, our framework exhibits steeper and more stable search trajectories. By enabling the efficient, automated composition of workflow with functional components, AgentSwift provides a scalable methodology to explore complex agent designs. Our framework serves as a launchpad for researchers to rapidly prototype and discover powerful agent architectures without the impediment of prohibitive evaluation costs.

Recent advancements in large language models (LLMs) have demonstrated remarkable text generation capabilities. However, controlling specific attributes of generated text remains challenging without architectural modifications or extensive fine-tuning. Current methods typically toggle a single, basic attribute but struggle with precise multi-attribute control. In scenarios where attribute requirements conflict, existing methods lack coordination mechanisms, causing interference between desired attributes. Furthermore, these methods fail to incorporate iterative optimization processes in the controlled generation pipeline. To address these limitations, we propose Conflict-aware, Composite, and Collaborative Controlled Text Generation (C³TG), a two-phase framework for fine-grained, multi-dimensional text attribute control. During generation, C³TG selectively pairs the LLM with the required attribute classifiers from the 17 available dimensions and employs weighted KL-divergence to adjust token probabilities. The optimization phase then leverages an energy function combining classifier scores and penalty terms to resolve attribute conflicts through iterative feedback, enabling precise control over multiple dimensions simultaneously while preserving natural text flow. Experiments show that C³TG significantly outperforms baselines across multiple metrics including attribute accuracy, linguistic fluency, and output diversity, while simultaneously reducing toxicity. These results establish C³TG as an effective and flexible solution for multi-dimensional text attribute control that requires no costly model modifications.

Backdoor attacks pose a severe threat to federated graph learning (FGL), where malicious clients can inject hidden triggers into the global model without being detected. Defending against such attacks is particularly challenging due to the complex graph structures and the stealthy nature of trigger patterns. In this work, we propose MultiKD, a novel backdoor mitigation method based on attention-guided multi-teacher distillation. Unlike existing defenses that focus on detecting suspicious clients or restricting backdoor activation, MultiKD directly purifies the global model on the server side by exploiting intermediate representations. It integrates knowledge from multiple client models and guides the global model to suppress backdoor behaviors by aligning attention maps and preserving inter-layer relational consistency. Our defensive intuition enables MultiKD to retain task-relevant information while mitigating malicious patterns, even when some teacher models are compromised. Extensive experiments on four real-world datasets demonstrate the effectiveness of our approach in significantly reducing attack success rate (≤ 8%) with minimal impact on utility (≤ 5%).

Multimodal recommendation, which integrates user-item interactions with rich item multimodal features to infer user preferences, has emerged as one of the state-of-the-art recommendation paradigms. However, most existing multimodal recommendation methods focus solely on the multimodal fusion process of items, overlooking the impact of modality-specific noise characteristics and inter-modal distribution shifts on the fusion process. As a result, noise from specific modalities may be inadvertently amplified during the fusion of multimodal representations, thereby degrading the quality of fused representations and ultimately undermining recommendation accuracy. To address this issue, we propose a novel framework, DAMPS, which Denoises and cAlibrates multimodal representations in the spectral domain from the perspectives of both aMplitude and Phase Spectrum. Specifically, we first apply the Fast Fourier Transform to decompose multimodal representations into amplitude spectrum and phase spectrum components. We then propose the adaptive phase calibrator, which resolves the phase distribution shift between modalities through the optimal minimum variance unbiased estimate of phase offset and adaptive phase rotation. We further develop the amplitude variance-ratio filter, which suppresses modality-specific amplitude noise while preserving prominent modality-specific representations. Finally, we introduce the inter-modal coherence filter, which quantifies and enhances the intrinsic shared information between modalities through amplitude-squared coherence, thereby eliminating irrelevant modality noise. Technically, DAMPS is model-agnostic and highly flexible, enabling seamless integration with various multimodal recommendation backbones. Extensive evaluations on five public multimodal recommendation datasets demonstrate that DAMPS consistently improves recommendation performance and achieves state-of-the-art results.

Diffusion models have achieved significant progress in both image and video generation while still suffering from huge computation costs. As an effective solution, rectified flow aims to rectify the diffusion process of diffusion models into a straight line for few-step and even one-step generation. However, in this paper, we suggest that the original training pipeline of reflow is not optimal and introduce two techniques to improve it. Firstly, we introduce progressive reflow, which progressively reflows the diffusion models in local timesteps until the whole diffusion progresses, reducing the difficulty of flow matching. Second, we introduce aligned v-prediction, which highlights the importance of direction matching in flow matching over magnitude matching. Experimental results on SDv1.5 and SDXL demonstrate the effectiveness of our method, for example, conducting on SDv1.5 achieves an FID of 10.70 on MSCOCO2014 validation set with only 4 sampling steps, close to our teacher model (32 DDIM steps, FID = 10.05). Our codes will be released at Github.

Articulated objects pose diverse manipulation challenges for robots. Since their internal structures are not directly observable, robots must adaptively explore and refine actions to generate successful manipulation trajectories. While existing works have attempted cross-category generalization in adaptive articulated object manipulation, two major challenges persist: (1) the geometric diversity of real-world articulated objects complicates visual perception and understanding, and (2) variations in object functions and mechanisms hinder the development of a unified adaptive manipulation strategy. To address these challenges, we propose AdaRPG, a novel framework that leverages foundation models to extract object parts, which exhibit greater local geometric similarity than entire objects, thereby enhancing visual affordance generalization for functional primitive skills. To support this, we construct a part-level affordance annotation dataset to train the affordance model. Additionally, AdaRPG utilizes the common knowledge embedded in foundation models to reason about complex mechanisms and generate high-level control codes that invoke primitive skill functions based on part affordance inference. Simulation and real-world experiments demonstrate AdaRPG's strong generalization ability across novel articulated object categories.

Recent advancements in Large Language Models (LLMs) have led to a rapid growth of agentic systems capable of handling a wide range of complex tasks. However, current research largely relies on manual, task-specific design, limiting their adaptability to novel tasks. In this paper, we introduce a new research problem: Modularized LLM Agent Search (MoLAS). We propose a modular design space that abstracts existing LLM agent designs into four fundamental modules with uniform IO interface: Planning, Reasoning, Tool Use, and Memory. Building on this design space, we present a novel LLM agent search framework called AgentSquare, which introduces two core mechanisms, i.e., module evolution and recombination, to efficiently search for optimized LLM agents. To further accelerate the process, we design a performance predictor that uses in-context surrogate models to skip unpromising agent designs. Extensive experiments across six benchmarks, covering the diverse scenarios of web, embodied, tool use and game applications, show that AgentSquare substantially outperforms hand-crafted agents, achieving an average performance gain of 17.2% against best-known human designs. Moreover, AgentSquare can generate interpretable design insights, enabling a deeper understanding of agentic architecture and its impact on task performance. We believe that the modular design space and AgentSquare search framework offer a platform for fully exploiting the potential of prior successful designs and consolidate the collective efforts of research community. Code repo is available at https://github.com/tsinghua-fib-lab/AgentSquare.

Articulated object manipulation is a critical capability for robots to perform various tasks in real-world scenarios. Composed of multiple parts connected by joints, articulated objects are endowed with diverse functional mechanisms through complex relative motions. For example, a safe consists of a door, a handle, and a lock, where the door can only be opened when the latch is unlocked. The internal structure, such as the state of a lock or joint angle constraints, cannot be directly observed from visual observation. Consequently, successful manipulation of these objects requires adaptive adjustment based on trial and error rather than a one-time visual inference. However, previous datasets and simulation environments for articulated objects have primarily focused on simple manipulation mechanisms where the complete manipulation process can be inferred from the object's appearance. To enhance the diversity and complexity of adaptive manipulation mechanisms, we build a novel articulated object manipulation environment and equip it with 9 categories of objects. Based on the environment and objects, we further propose an adaptive demonstration collection and 3D visual diffusion-based imitation learning pipeline that learns the adaptive manipulation policy. The effectiveness of our designs and proposed method is validated through both simulation and real-world experiments.

3D articulated objects are inherently challenging for manipulation due to the varied geometries and intricate functionalities associated with articulated objects. Point-level affordance, which predicts the per-point actionable score and thus proposes the best point to interact with, has demonstrated excellent performance and generalization capabilities in articulated object manipulation. However, a significant challenge remains: while previous works use perfect point cloud generated in simulation, the models cannot directly apply to the noisy point cloud in the real-world. To tackle this challenge, we leverage the property of real-world scanned point cloud that, the point cloud becomes less noisy when the camera is closer to the object. Therefore, we propose a novel coarse-to-fine affordance learning pipeline to mitigate the effect of point cloud noise in two stages. In the first stage, we learn the affordance on the noisy far point cloud which includes the whole object to propose the approximated place to manipulate. Then, we move the camera in front of the approximated place, scan a less noisy point cloud containing precise local geometries for manipulation, and learn affordance on such point cloud to propose fine-grained final actions. The proposed method is thoroughly evaluated both using large-scale simulated noisy point clouds mimicking real-world scans, and in the real world scenarios, with superiority over existing methods, demonstrating the effectiveness in tackling the noisy real-world point cloud problem.

This paper studies learning meaningful node representations for signed graphs, where both positive and negative links exist. This problem has been widely studied by meticulously designing expressive signed graph neural networks, as well as capturing the structural information of the signed graph through traditional structure decomposition methods, e.g., spectral graph theory. In this paper, we propose a novel signed graph representation learning framework, called Signed Laplacian Graph Neural Network (SLGNN), which combines the advantages of both. Specifically, based on spectral graph theory and graph signal processing, we first design different low-pass and high-pass graph convolution filters to extract low-frequency and high-frequency information on positive and negative links, respectively, and then combine them into a unified message passing framework. To effectively model signed graphs, we further propose a self-gating mechanism to estimate the impacts of low-frequency and high-frequency information during message passing. We mathematically establish the relationship between the aggregation process in SLGNN and signed Laplacian regularization in signed graphs, and theoretically analyze the expressiveness of SLGNN. Experimental results demonstrate that SLGNN outperforms various competitive baselines and achieves state-of-the-art performance.

Graph neural networks (GNNs) have received tremendous attention due to their superiority in learning node representations. These models rely on message passing and feature transformation functions to encode the structural and feature information from neighbors. However, stacking more convolutional layers significantly decreases the performance of GNNs. Most recent studies attribute this limitation to the over-smoothing issue, where node embeddings converge to indistinguishable vectors. Through a number of experimental observations, we argue that the main factor degrading the performance is the unstable forward normalization and backward gradient resulted from the improper design of the feature transformation, especially for shallow GNNs where the over-smoothing has not happened. Therefore, we propose a novel orthogonal feature transformation, named Ortho-GConv, which could generally augment the existing GNN backbones to stabilize the model training and improve the model's generalization performance. Specifically, we maintain the orthogonality of the feature transformation comprehensively from three perspectives, namely hybrid weight initialization, orthogonal transformation, and orthogonal regularization. By equipping the existing GNNs (e.g. GCN, JKNet, GCNII) with Ortho-GConv, we demonstrate the generality of the orthogonal feature transformation to enable stable training, and show its effectiveness for node and graph classification tasks.

Learning the low-dimensional representations of graphs (i.e., network embedding) plays a critical role in network analysis and facilitates many downstream tasks. Recently graph convolutional networks (GCNs) have revolutionized the field of network embedding, and led to state-of-the-art performance in network analysis tasks such as link prediction and node classification. Nevertheless, most of the existing GCN-based network embedding methods are proposed for unsigned networks. However, in the real world, some of the networks are signed, where the links are annotated with different polarities, e.g., positive vs. negative. Since negative links may have different properties from the positive ones and can also significantly affect the quality of network embedding. Thus in this paper, we propose a novel network embedding framework SNEA to learn Signed Network Embedding via graph Attention. In particular, we propose a masked self-attentional layer, which leverages self-attention mechanism to estimate the importance coefficient for pair of nodes connected by different type of links during the embedding aggregation process. Then SNEA utilizes the masked self-attentional layers to aggregate more important information from neighboring nodes to generate the node embeddings based on balance theory. Experimental results demonstrate the effectiveness of the proposed framework through signed link prediction task on several real-world signed network datasets.

Network embedding is an effective approach to learn the low-dimensional representations of vertices in networks, aiming to capture and preserve the structure and inherent properties of networks. The vast majority of existing network embedding methods exclusively focus on vertex proximity of networks, while ignoring the network internal community structure. However, the homophily principle indicates that vertices within the same community are more similar to each other than those from different communities, thus vertices within the same community should have similar vertex representations. Motivated by this, we propose a novel network embedding framework NECS to learn the Network Embedding with Community Structural information, which preserves the high-order proximity and incorporates the community structure in vertex representation learning. We formulate the problem into a principled optimization framework and provide an effective alternating algorithm to solve it. Extensive experimental results on several benchmark network datasets demonstrate the effectiveness of the proposed framework in various network analysis tasks including network reconstruction, link prediction and vertex classification.