Deepfake detection is a long-established research topic vital for mitigating the spread of malicious misinformation. Unlike prior methods that provide either binary classification results or textual explanations separately, we introduce a novel method capable of generating both simultaneously. Our method harnesses the multi-modal learning capability of the pre-trained CLIP and the unprecedented interpretability of large language models (LLMs) to enhance both the generalization and explainability of deep-fake detection. Specifically, we introduce a multi-modal face forgery detector (M2F2-Det) that employs tailored face forgery prompt learning, incorporating the pre-trained CLIP to improve generalization to unseen forgeries. Also, M2F2-Det incorporates an LLM to provide detailed textual explanations of its detection decisions, enhancing interpretability by bridging the gap between natural language and subtle cues of facial forgeries. Empirically, we evaluate M2F2-Det on both detection and explanation generation tasks, where it achieves state-of-the-art performance, demonstrating its effectiveness in identifying and explaining diverse forgeries. Source code is available at $\color{magenta}{link}$.

Traditional spatiotemporal models generally rely on task-specific architectures, which limit their generalizability and scalability across diverse tasks due to domain-specific design requirements. In this paper, we introduce UniSTD, a unified Transformer-based framework for spatiotemporal modeling, which is inspired by advances in recent foundation models with the two-stage pretraining-then-adaption paradigm. Specifically, our work demonstrates that task-agnostic pretraining on 2D vision and vision-text datasets can build a generalizable model foundation for spatiotemporal learning, followed by specialized joint training on spatiotemporal datasets to enhance task-specific adaptability. To improve the learning capabilities across domains, our framework employs a rank-adaptive mixture-of-expert adaptation by using fractional interpolation to relax the discrete variables so that can be optimized in the continuous space. Additionally, we introduce a temporal module to incorporate temporal dynamics explicitly. We evaluate our approach on a large-scale dataset covering 10 tasks across 4 disciplines, demonstrating that a unified spatiotemporal model can achieve scalable, cross-task learning and support up to 10 tasks simultaneously within one model while reducing training costs in multi-domain applications. Code will be available at https://github.com/1hunters/UniSTD.

Designing effective embodied multi-agent systems is critical for solving complex real-world tasks across domains. Due to the complexity of multi-agent embodied systems, existing methods fail to automatically generate safe and efficient training data for such systems. To this end, we propose the concept of compositional constraints for embodied multi-agent systems, addressing the challenges arising from collaboration among embodied agents. We design various interfaces tailored to different types of constraints, enabling seamless interaction with the physical world. Leveraging compositional constraints and specifically designed interfaces, we develop an automated data collection framework for embodied multi-agent systems and introduce the first benchmark for embodied multi-agent manipulation, RoboFactory. Based on RoboFactory benchmark, we adapt and evaluate the method of imitation learning and analyzed its performance in different difficulty agent tasks. Furthermore, we explore the architectures and training strategies for multi-agent imitation learning, aiming to build safe and efficient embodied multi-agent systems.

Coordinating multiple embodied agents in dynamic environments remains a core challenge in artificial intelligence, requiring both perception-driven reasoning and scalable cooperation strategies. While recent works have leveraged large language models (LLMs) for multi-agent planning, a few have begun to explore vision-language models (VLMs) for visual reasoning. However, these VLM-based approaches remain limited in their support for diverse embodiment types. In this work, we introduce VIKI-Bench, the first hierarchical benchmark tailored for embodied multi-agent cooperation, featuring three structured levels: agent activation, task planning, and trajectory perception. VIKI-Bench includes diverse robot embodiments, multi-view visual observations, and structured supervision signals to evaluate reasoning grounded in visual inputs. To demonstrate the utility of VIKI-Bench, we propose VIKI-R, a two-stage framework that fine-tunes a pretrained vision-language model (VLM) using Chain-of-Thought annotated demonstrations, followed by reinforcement learning under multi-level reward signals. Our extensive experiments show that VIKI-R significantly outperforms baselines method across all task levels. Furthermore, we show that reinforcement learning enables the emergence of compositional cooperation patterns among heterogeneous agents. Together, VIKI-Bench and VIKI-R offer a unified testbed and method for advancing multi-agent, visual-driven cooperation in embodied AI systems.

Large numbers of synthesized videos from diffusion models pose threats to information security and authenticity, leading to an increasing demand for generated content detection. However, existing video-level detection algorithms primarily focus on detecting facial forgeries and often fail to identify diffusion-generated content with a diverse range of semantics. To advance the field of video forensics, we propose an innovative algorithm named Multi-Modal Detection(MM-Det) for detecting diffusion-generated videos. MM-Det utilizes the profound perceptual and comprehensive abilities of Large Multi-modal Models (LMMs) by generating a Multi-Modal Forgery Representation (MMFR) from LMM's multi-modal space, enhancing its ability to detect unseen forgery content. Besides, MM-Det leverages an In-and-Across Frame Attention (IAFA) mechanism for feature augmentation in the spatio-temporal domain. A dynamic fusion strategy helps refine forgery representations for the fusion. Moreover, we construct a comprehensive diffusion video dataset, called Diffusion Video Forensics (DVF), across a wide range of forgery videos. MM-Det achieves state-of-the-art performance in DVF, demonstrating the effectiveness of our algorithm. Both source code and DVF are available at https://github.com/SparkleXFantasy/MM-Det.

The amplitude‐and‐phase error (APE) between phased array channels is notorious in radar signal processing. This error can cause an inaccurate estimate of unknown steering vector of the target echo signal and eventually result in amplitude‐phase discontinuity of the phased array output. Thus, how to handle the APE is a meaningful problem, particularly for the varying jamming environment, of which the signal‐to‐noise ratio is very low. In this study, the authors have developed a new method to obtain real‐time amplitude and phase differences between two consecutive weight update periods based on the accurate estimation of steering vector. Such differences can be used to obtain a real‐time weight vector with negligible amplitude‐phase distortions, and hence improves the phased array signal‐processing performance. The proposed method is very flexible: it works well in different array configurations, such as linear, rectangular and Y‐shape arrays, and can be efficiently implemented in any eigenstructure‐based direction‐of‐arrival system.

A voltage/current transient typically caused by islanding and switching operations is treated as an adverse phenomenon that degrades power quality, and it may cause damage to electrical equipment. Therefore, a reliable system should effectively detect and monitor a transient disturbance. In this paper, the transient detection problem is formulated as a binary hypothesis test: normal signal (null) vs. transient (alternative). The sampled data is described by a sinusoid under the null hypothesis, while a sum of damped sinusoids is utilized to model the alternative one. As no prior knowledge is imposed on complex amplitudes, frequencies, or damping factors in signal modeling, the general likelihood ratio test (GLRT) is employed to fulfill the task. To reduce computational complexity, the maximum likelihood estimator is replaced by ESPRIT for parameter estimation. Probability of detection of 0.98 is achieved at a SNR of 27dB and probability of false alarm of 0.0005.

This paper investigates moving target localization by the biased time-of-arrivals (TOAs), where an extracted TOA is biased by the unknown Doppler of the target. The phenomenon applies to radar and sonar systems that employ Doppler tolerant waveforms. While in principle those would allow the estimation of both position and velocity using position-only measurements, we unfortunately find that the Fisher information matrix is ill-conditioned. However, a Quasi-maximum likelihood estimator based on the misspecified model is suggested, and its performance is analyzed.

A multiple-input multiple-output (MIMO) radar can improve system performance with waveform and spatial diversities. Mathematically, the multiple independent waveforms increase the dimension of signal space, so optimal transmission power allocation deserves investigation. The majority of the literature prefers to omit the effect of propagation attenuation, and considers the receiving gain vectors to be independently and identically distributed (i.i.d.) in power allocation. In this paper, we integrate the propagation losses into MIMO radar signal model, and investigate the power allocation problems under three popular criteria: maximizing the mutual information, minimizing the minimum mean square errors, and maximizing the echo energy. As their objective functions are either convex or concave, the optimal strategies are theoretically guaranteed.

The interaction between a smart target and a smart MIMO radar is investigated from a game theory perspective. Since the target and the radar form an adversarial system, their interaction is modeled as a two-person zero-sum game. The mutual information criterion is used to formulate the utility functions. The unilateral, hierarchical, and symmetric games are studied, and the equilibria solutions are derived.

Bearings-only localization with light-of-sight (LOS) propagation is well understood. This paper concentrates on bearing-only localization with non-line-of-sight (NLOS) measurements, where the target radiation arrives at the sensor after a specular reflection. The wrinkle is that the reflecting surface is circular (inner side of a circle), and is assumed known. Since the target-sensor geometry has multiple configurations, the maximum likelihood (ML) solution may not exist. However, if a concentric opaque circle (such as the earth) exists within the reflecting one, the propagation path is unique; a grid search based ML is available for such a circumstance. Since ML is computationally consuming, two suboptimal algorithms based on small angle approximation are developed. Their performances are numerically compared.

As is clear from the ambiguity function (AF) uncertainty principle and the underlying matched filtering operation, pulse compression (PC) radars cannot accurately and simultaneously measure both the time delay and Doppler shift of moving targets. On the other hand, compressed sensing (CS) radar would seem to be emerging as a means to do just that, jointly to estimate the delay and Doppler shift, through convex optimization instead of matched filtering, such that PC resolution limits no longer apply. However, it turns out that the AF is closely related to the CS “dictionary coherence coefficient,” which affects the optimization recoverability. In this paper we formalize this relationship.