Current fake image detectors trained on large synthetic image datasets perform satisfactorily on limited studied generative models. However, these detectors suffer a notable performance decline over unseen models. Besides, collecting adequate training data from online generative models is often expensive or infeasible. To overcome these issues, we propose Few-Shot Detector (FSD), a novel AI-generated image detector which learns a specialized metric space for effectively distinguishing unseen fake images using very few samples. Experiments show that FSD achieves state-of-the-art performance by $+11.6\%$ average accuracy on the GenImage dataset with only $10$ additional samples. More importantly, our method is better capable of capturing the intra-category commonality in unseen images without further training. Our code is available at https://github.com/teheperinko541/Few-Shot-AIGI-Detector.

Group role assignment (GRA) maximizes total benefits by assigning agents to appropriate roles, while GRA with a training plan (GRATP) further considers the impact of training. However, existing research on GRA and GRATP does not fully consider the demand for flexible adjustment of human resource assignment, which may lead to increased employment costs and project delays. Moreover, role importance significantly affects training resource assignment, as key roles contribute more to overall performance. Therefore, we propose the GRA with secondment and role-importance-based training plan (GRA-SRIT) model to address these issues. Specifically, this article introduces seconded personnel to temporarily replace the positions of agents undergoing training, ensuring the smooth continuation of the project. Depending on the role requirements, different training durations are assigned based on the specific requirements of their roles. In addition, trainers with different levels of expertise are assigned to agents based on role importance, ensuring that critical roles receive more specialized training, thus maximizing total benefit. Finally, experiments demonstrate the proposed model’s effectiveness in different scenarios.

Role-based collaboration (RBC) is an innovative collaborative approach designed to enhance collaboration. Role negotiation (RN) is a critical step in RBC, during which the role set and the number of agents required for each role, i.e., role requirements, are determined. This process establishes the foundational input for group role assignment (GRA), where roles are assigned to agents to optimize group performance. Uncertainties in RN, such as task volume fluctuations, create dynamic agent requirements. However, existing RBC models typically assume RN to be static, thus failing to adequately address the substantial challenges. Three-way decision (3WD) is a robust decision-making methodology well-suited for managing uncertainty. To address the uncertainties in role requirements, this article introduces truncated discrete distribution to quantify role requirements, and presents a novel group role three-way assignment (GR3A) model. Compared with traditional RBC, our model offers an additional variable partial substitute choice that offers agents little salary during nonengagement periods but can transition to full involvement as required according to the prior agreement. GR3A is a dual-objective nonlinear optimization problem, for which a linearization strategy is proposed to achieve the optimal resolution. Additionally, sufficient and necessary conditions for these assignment problems are put forward to enhance the efficacy of the proposed solutions. To our knowledge, this study innovatively introduces a truncated discrete distribution and 3WD into the RBC framework. Empirical validation through simulations demonstrates the effectiveness and efficacy of the proposed method within the RBC context.

Role-based collaboration (RBC) is a novel problem-solving paradigm to facilitate collaboration. Group multirole assignment (GMRA), an extension of group role assignment (GRA), is a critical step in the RBC process, enabling the formation of efficient collaborative teams by reasonably assigning roles to agents. Recognized as a significant determinant impacting assignment and collaboration, conflict has been delineated and incorporated into GMRA. Nevertheless, the specified conflict is characterized as an oppositional conflict (OC), signifying that conflicting agents are engaged in conflict across the entirety of the role set. Rather than completely OC, which is highly specific, a more prevalent relationship involves conflict in certain aspects while remaining conflict-free in others. Therefore, we propose the concept of general conflict (GC) to describe the more common and realistic conflict relationship, offering a broader and novel perspective to depict conflict relationships. Then, we formalize these two problems by considering GC avoidance in GRA and GMRA, called GRA with GC (GRAGC) and GMRA with GC (GMRAGC), respectively. Furthermore, we establish the necessary conditions for the GRAGC and GMRAGC problems through a graph-theoretical lens, accompanied by a thorough analysis of their mathematical nature. Additionally, we propose practical solutions and refine methodologies to address both problems. The effectiveness of the improved algorithms utilizing necessary conditions is verified by simulations, which also provides evidence supporting the advantages of conflict avoidance.

Spatio-temporal information plays an important role in compressed video quality enhancement. Most advanced studies use deformable convolution or Swin transformer to explore spatio-temporal information. However, deformable convolution based methods may incur inaccurate motion compensation due to the compression artifacts and limited receptive fields. The Swin transformer based approaches are unable to fully explore the spatio-temporal information, limited by its rigid window-based mechanism. To solve the above problems, we propose a novel multi-Swin transformer-based network for compressed video quality enhancement to better explore spatio-temporal information. The whole workflow consists of the Local Alignment (LA) Module, the Global Refinement Fusion (GRF) Module, and the Quality Enhancement (QE) Module. The LA module roughly perceives the local motion through the deformable fusion. Subsequently, the GRF module employs the proposed multi-Swin transformer to enhance the spatio-temporal perception. Finally, the QE module effectively restores the texture details across various scales. Extensive experimental results prove the effectiveness of the proposed method.

Based on role-based collaboration (RBC), group role assignment (GRA) optimizes a team’s overall performance by assigning the most appropriate individual agents from the team’s viewpoint based on agents’ role-playing abilities. As an extension of GRA, GRA with a training plan (GRATP) deals with the impact of training on team management. Considering the correlation between roles, the training of one agent on one role also affects the performance of the agent in other roles. Moreover, in the adaptive collaboration (AC) problem, the training time also affects significantly the agent’s ability, as an agent’s ability changes over time. However, the existing GRATP models fail to consider these factors in the collaboration process. Therefore, we aim to address the role-correlation-based adaptive GRATP (RCA-GRATP) in this article. This article contributes two aspects to the literature on AC. 1) RCA-GRATP problem is abstracted based on RBC and GRA. To the best of the authors’ knowledge, this is the first article that explicitly considers role correlation in the RBC problems. 2) A comprehensive formalization of RCA-GRATP and two solving algorithms for diverse situations are proposed to solve the formalized problems. Experiments are carried out to verify the effectiveness of the proposed algorithms in diverse scenarios.

Group Role Assignment (GRA) seeks the maximum group performance by assigning roles to the most appropriate individual agents. As an extension of GRA, GRA with a Training Plan (GRATP) is proposed to address the trainingrelated collaboration problems. However, the existing models neglect the influence of changes in duration, i.e., the duration of training is a preset constant, which has limited applications. The variation of duration has a great influence on the improvement of the agent’s ability and assignment, especially in the dynamic scenario. Therefore, this paper formalizes a GRATP problem that considers the duration in dynamic scenarios, and constructs an algorithm to solve it. In the formulated problem, the training plan includes three factors: the starting time, the duration and the training programs. The total benefit is taken as the goal of the algorithm, rather than the group performance, because the agent’s ability and group performance vary over time in adaptive collaboration (AC). Moreover, the reassignment needs to be launched after training, as the original assignment may not be optimal. By utilizing GRA and its Environment—Classes, Agents, Roles, Groups, and Objects (E-CARGO) model, the optimal training plan and reassignment are obtained by maximizing the total benefit. Experiments verify the effectiveness of the proposed algorithm.

This paper focuses on energy efficiency (EE) pre-coding design and computational-efficient precoding updating strategy for low earth orbit (LEO) satellite communications. Firstly, we formulate the EE precoding problem, which aims to maximize the EE metric under the quality of service (QoS) constraint and per-antenna power constraint (PAPC). By intro-ducing semidefinite relaxation, first-order Taylor approximation, and quadratic transformation, the problem is transferred into a convex one that can be efficiently solved. Moreover, due to the continuous movement of LEO satellites, precoding is performed frequently to maintain the high EE performance, leading to high computational complexity. Consequently, we consider prolonging precoding intervals to reduce complexity while alleviating severe performance degradation during the intervals. To this end, a computational-efficient beam direction change (BDC) algorithm is proposed to update pre coding vectors, which makes the main lobes of beams always point toward users. Furthermore, an adaptive method is proposed to adjust the precoding interval flexibly. Simulation results have indicated the effectiveness of the EE precoding algorithm and the BDC algorithm.

Abstract With the increasing number of user terminals (UTs), the interference among UTs might significantly decrease the throughput of the low earth orbit satellite communication system. In this paper, the user scheduling method is investigated to suppress user interference. Specifically, leveraging the strong spatial directivity of satellite channels, a low‐complexity angle‐based orthogonal user selection (AOUS) algorithm is proposed, which selects UTs with nearly orthogonal channels via angle information of UTs. A rate‐based proportionally fair (PF)‐AOUS algorithm is further proposed to ensure fairness among UTs, which combines the AOUS with the PF criterion. To reduce complexity, an improved angle‐based PF‐AOUS algorithm that schedules UTs according to their pitch angles rather than their rates is proposed. In addition, efficient precoding schemes for orthogonal UTs are designed by combining the steering vector and power allocation matrix, and it is shown that precoding can be converted into power allocation problems that further balance fairness and throughput. The numerical results indicate that the AOUS achieves a near‐optimal sum rate performance, and the angle‐based PF‐AOUS has the similar performance to the rate‐based PF‐AOUS, which achieves a high fairness index with the proposed precoding scheme.

The three-way decision (3WD) using probabilistic rough set (PRS) is constructed based on the probabilistic thresholds, and the 3WD using decision-theoretic rough set (DTRS) can utilize the decision cost information. They are both popular and effective decision-making methods. In the same problem, experts may provide different types of information about the decision rules. However, few works try to combine these two kinds of models and utilize both kinds of information. Therefore, in this paper, we propose a 3WD model combining PRS and DTRS, which integrates the probabilistic risk and the Bayesian risk in decision problems. To achieve that, based on the given probabilistic thresholds and decision rules, the 3WD with PRS is analyzed from the perspective of decision risk. Following that, the total risk of three decisions is computed. Then the best option for the instance is selected according to the Bayesian minimum risk rule. After the existence and uniqueness of the thresholds are analyzed, the explicit expressions are provided. Finally, illustration examples demonstrate the effectiveness of the proposed approach.

Training is an effective way to improve agents’ performance. As an extension of group role assignment (GRA), GRA with a training plan (GRATP) aims to maximize the group performance or benefit by finding the optimal role assignment and training plan. However, GRATP has only been discussed in static scenarios. In dynamic environments, agents’ performance changes over time and adaptive collaboration is designed to keep the group in a good state. Therefore, this paper investigates the GRATP problem in adaptive collaboration. The timing of training has a significant impact on the improvement of individual performance, which will in turn affect the team performance and total benefit. By utilizing Role-Based Collaboration and GRA, the optimal training timing and training plan are obtained, which maximize the total benefit of the group. After training, the roles are re-assigned to the agents based on their current performance. This paper’s contributions include formalizing the GRATP problem in adaptive collaboration and presenting a solution to it. The effectiveness of the proposed method is verified by experiments.

Aiming at the problems of low resolution and poor imaging quality in global satellite system reflectometry based synthetic aperture radar (GNSS-R SAR), this paper proposes a multi-period SAR imaging method with a GPS satellite and a fixed receiver. Coherent fusion will be used in point target simulation. The result achieves the improvement of imaging resolution. In the actual scene experiment, with the larger interval of the data acquisition time, finally, the more abundant image feature information is obtained compared with the traditional bistatic SAR imaging. This method proves the vast potential in realizing local area monitoring by GNSS-R SAR.

Robotic Learning from Demonstration (LfD) allows anyone, not just experts, to program a robot for an arbitrary task. Many LfD methods focus on low level primitive actions such as manipulator trajectories. Complex multistep task with many primitive actions must be learned from demonstration if LfD is to encompass the full range of task a user may desire. Existing methods represent the high level task in various forms including, finite state machines, decision trees, formal logic, among others. Behavior trees are proposed as an alternative representation of high level task. Behavior trees are an execution model for the control of a robot designed for real time execution, modularity, and, consequently, transparency. Real time execution allows the robot to reactively perform the task. Modularity allows the reuse of learned primitive actions and high level task in new situations, speeding up the process of learning in new scenarios. Transparency allows users to understand and interactively modify the learned model. Behavior trees are used to represent high level tasks by building on the relationship it has with decision trees. We demonstrate a human teaching our Fetch robot a household cleaning task.

Transparent objects are prevalent across many environments of interest for dexterous robotic manipulation. Such transparent material leads to considerable uncertainty for robot perception and manipulation, and remains an open challenge for robotics. This problem is exacerbated when multiple transparent objects cluster into piles of clutter. In household environments, for example, it is common to encounter piles of glassware in kitchens, dining rooms, and reception areas, which are essentially invisible to modern robots. We present the GlassLoc algorithm for grasp pose detection of transparent objects in transparent clutter using plenoptic sensing. GlassLoc classifies graspable locations in space informed by a Depth Likelihood Volume (DLV) descriptor. We extend the DLV to infer the occupancy of transparent objects over a given space from multiple plenoptic viewpoints. We demonstrate and evaluate the GlassLoc algorithm on a Michigan Progress Fetch mounted with a first generation Lytro. The effectiveness of our algorithm is evaluated through experiments for grasp detection and execution with a variety of transparent glassware in minor clutter.

Recently, neural network based sentence modeling methods have achieved great progress.Among these methods, the recursive neural networks (RecNNs) can effectively model the combination of the words in sentence.However, RecNNs need a given external topological structure, like syntactic tree.In this paper, we propose a gated recursive neural network (GRNN) to model sentences, which employs a full binary tree (FBT) structure to control the combinations in recursive structure.By introducing two kinds of gates, our model can better model the complicated combinations of features.Experiments on three text classification datasets show the effectiveness of our model.

Neural network based methods have obtained great progress on a variety of natural language processing tasks.However, it is still a challenge task to model long texts, such as sentences and documents.In this paper, we propose a multi-timescale long short-term memory (MT-LSTM) neural network to model long texts.MT-LSTM partitions the hidden states of the standard LSTM into several groups.Each group is activated at different time periods.Thus, MT-LSTM can model very long documents as well as short sentences.Experiments on four benchmark datasets show that our model outperforms the other neural models in text classification task.

In this paper, we give an overview for the shared task at the 4th CCF Conference on Natural Language Processing & Chinese Computing (NLPCC 2015): Chinese word segmentation and part-of-speech (POS) tagging for micro-blog texts. Different with the popular used newswire datasets, the dataset of this shared task consists of the relatively informal micro-texts. The shared task has two sub-tasks: (1) individual Chinese word segmentation and (2) joint Chinese word segmentation and POS Tagging. Each subtask has three tracks to distinguish the systems with different resources. We first introduce the dataset and task, then we characterize the different approaches of the participating systems, report the test results, and provide a overview analysis of these results. An online system is available for open registration and evaluation at http://nlp.fudan.edu.cn/nlpcc2015 .