Human-centric perceptions play a crucial role in real-world applications. While recent human-centric works have achieved impressive progress, these efforts are often constrained to the visual domain and lack interaction with human instructions, limiting their applicability in broader scenarios such as chatbots and sports analysis. This paper introducesReferring Human Perceptions, where a referring prompt specifies the person of interest in an image. To tackle the new task, we propose RefHCM (ReferringHuman-CentricModel), a unified framework to integrate a wide range of human-centric referring tasks. Specifically, RefHCM employs sequence mergers to convert raw multimodal data—including images, text, coordinates, and parsing maps—into semantic tokens. This standardized representation enables RefHCM to reformulate diverse human-centric referring tasks into a sequence-to-sequence paradigm, solved using a plain encoder-decoder transformer architecture. Benefiting from a unified learning strategy, RefHCM effectively facilitates knowledge transfer across tasks and exhibits unforeseen capabilities in handling complex reasoning. This work represents the first attempt to address referring human perceptions with a general-purpose framework, while simultaneously establishing a corresponding benchmark that sets new standards for the field. Extensive experiments showcase RefHCM's competitive and even superior performance across multiple human-centric referring tasks. The code and data are publicly athttps://github.com/JJJYmmm/RefHCM.

In video Multimodal Large Language Models (video MLLMs), the visual encapsulation process plays a pivotal role in converting video contents into representative tokens for LLM input. While linear projectors are widely employed for encapsulation, they introduce semantic indistinctness and temporal incoherence when applied to videos. Conversely, the structure of resamplers shows promise in tackling these challenges, but an effective solution remains unexplored. Drawing inspiration from resampler structures, we introduce DisCo, a novel visual encapsulation method designed to yield semantically distinct and temporally coherent visual tokens for video MLLMs. DisCo integrates two key components: (1) A Visual Concept Discriminator (VCD) module, assigning unique semantics for visual tokens by associating them in pair with discriminative concepts in the video. (2) A Temporal Focus Calibrator (TFC) module, ensuring consistent temporal focus of visual tokens to video elements across every video frame. Through extensive experiments on multiple video MLLM frameworks, we demonstrate that DisCo remarkably outperforms previous state-of-the-art methods across a variety of video understanding benchmarks, while also achieving higher token efficiency thanks to the reduction of semantic indistinctness. The code: https://github.com/ZJHTerry18/DisCo.

Contrastive Language-Image Pre-training (CLIP) has become a foundation model and has been applied to various vision and multimodal tasks. However, recent works indicate that CLIP falls short in distinguishing detailed differences in images and shows suboptimal performance on dense-prediction and vision-centric multimodal tasks. Therefore, this work focuses on improving existing CLIP models, aiming to capture as many visual details in images as possible. We find that a specific type of generative models, unCLIP, provides a suitable framework for achieving our goal. Specifically, unCLIP trains an image generator conditioned on the CLIP image embedding. In other words, it inverts the CLIP image encoder. Compared to discriminative models like CLIP, generative models are better at capturing image details because they are trained to learn the data distribution of images. Additionally, the conditional input space of unCLIP aligns with CLIP's original image-text embedding space. Therefore, we propose to invert unCLIP (dubbed un$^2$CLIP) to improve the CLIP model. In this way, the improved image encoder can gain unCLIP's visual detail capturing ability while preserving its alignment with the original text encoder simultaneously. We evaluate our improved CLIP across various tasks to which CLIP has been applied, including the challenging MMVP-VLM benchmark, the dense-prediction open-vocabulary segmentation task, and multimodal large language model tasks. Experiments show that un$^2$CLIP significantly improves the original CLIP and previous CLIP improvement methods. Code and models are available at https://github.com/LiYinqi/un2CLIP.

Current clothes-changing person re-identification (re-id) approaches usually perform retrieval based on clothes-irrelevant features, while neglecting the potential of clothes-relevant features. However, we observe that relying solely on clothes-irrelevant features for clothes-changing re-id is limited, since they often lack adequate identity information and suffer from large intra-class variations. On the contrary, clothes-relevant features can be used to discover same-clothes intermediaries that possess informative identity clues. Based on this observation, we propose a Feasibility-Aware Intermediary Matching (FAIM) framework to additionally utilizeclothes-relevant featuresfor retrieval. First, an Intermediary Matching (IM) module is designed to perform an intermediary-assisted matching process. This process involves using clothes-relevant features to find informative intermediates, and then using clothes-irrelevant features of these intermediates to complete the matching. Second, in order to reduce the negative effect of low-quality intermediaries, an Intermediary-Based Feasibility Weighting (IBFW) module is designed to evaluate the feasibility of intermediary matching process by assessing the quality of intermediaries. Extensive experiments demonstrate that our method outperforms state-of-the-art methods on several widely-used clothes-changing re-id benchmarks.

Optical Network-on-Chip (ONoC) is a promising type of interconnection technology for high performance computing systems, with high bandwidth and low latency. In Optical Network-on-Chip (ONoC), accurate traffic prediction facilitates the path reservation and resource allocation in advance, which in turn reduces waiting delay and power overhead. In this work, neural network-based traffic prediction is proposed for ONoC, based on traffic matrix (TM). Different neural network-based traffic predictions are compared. In addition, inter-flow correlations are considered in the traffic prediction. The evaluation results show that LSTM-based traffic prediction performs more efficient. Also, the prediction results are more accurate incorporating inter-flow correlations.

As mission-critical Internet of Things (MC-IoT) is expected to carry important and private information, its high quality of service (QoS) and high physical layer (PHY) security are indispensable. Nevertheless, most existing PHY security related work is built on the assumption of infinite blocklength, which is not applicable to finite blocklength (FBL) transmission, a typical scenario in MC-IoT such as factory automation. In this paper, we address the PHY security issue of a hybrid duplex relay aided MC-IoT system with FBL. Closed-form expressions for statistical secrecy throughput of full-duplex (FD) and half-duplex (HD) relay systems are derived, respectively, which are verified by numerical results. Based on the closed-form secrecy throughput, joint optimization of blocklength and transmission powers at source and relay is conducted for FD and HD relay systems, respectively. A hybrid duplex relaying scheme is also proposed by selecting the duplex mode with a higher achievable secrecy throughput. Numerical results show that, together with the hybrid relaying scheme, the proposed relay system with joint power allocation and blocklength adaptation, relay mode selection achieves much higher secrecy throughput over the conventional sole FD or HD mode relaying systems. Also, it is revealed that increasing blocklength or transmitting power may not always lead to a higher secrecy throughput and energy efficiency (EE).

An interference alignment (IA) and independent component analysis (ICA)-based semiblind scheme, referred to as IA-ICA, is proposed for downlink dual-user power-domain nonorthogonal multiple access (NOMA) systems in high-reliability low-latency (HRLL) Internet of Things (IoT). At the base station (BS), one user is converted to constructive interference to the other user via phase alignment of each symbol. At both user ends, ICA is used for semiblind signal detection. The phase rotation via nonredundant precoding at the BS does not introduce any spectral overhead, while only 1-2 pilot symbols are required for elimination of ICA incurred ambiguity. Closed-form expressions are derived for the users' symbol error rate (SER) performance in Rayleigh fading with 4-quadrature amplitude modulation (4-QAM), which matches the simulation results very well. Based on the analytical results, we propose an efficient power allocation algorithm that is based on statistical channel state information (CSI) only, and therefore, the signaling overhead involved is negligible. In particular, a near-optimal SER performance can be achieved with equal power allocation between the two users. The proposed IA-ICA-based semiblind NOMA system demonstrates a much better SER performance than the existing approaches even though they are under perfect CSI. Hence, it is a feasible solution for HRLL IoT, with high reliability and very low spectral and signaling overheads.