Vision-Language Models (VLMs) have achieved impressive performance across various tasks, but often struggle to apply newly introduced visual concepts during inference. A common failure pattern is what we call Mixing Things Up: VLMs frequently confuse concept names, resulting in vague descriptions and failure to ground the concept correctly. Existing approaches mainly address person-related concepts through text prompts or tokenizer modifications. However, VLMs still miss or misinterpret untrained visual concepts, underscoring the need to learn new concepts directly from visual input, without relying on prior textual injection. To overcome these limitations, we propose BISCUIT (Basis-aligned Inference through Structured Concept Unification and Identification-aware Tuning), a two-step training method. Step I proposes a dual-stream structure-aware vision encoder that fuses RGB and edge-based embeddings within a shared basis space to enhance concept recognition. Step II enhances generation quality through identification-aware tuning, which encourages alignment between the generated text and the newly introduced visual concepts. Existing methods mainly focus on person concepts and lack comprehensive evaluation across diverse visual categories. We further propose a benchmark BiscuitVQA to evaluate VLMs performance on recognizing and applying novel image-introduced concepts across diverse concept types and task types, including real people, cartoons, animals, and symbolic content. We apply BISCUIT to LLaVA-1.5 and Qwen2.5-VL, achieving competitive results among open-source models and narrowing the gap to Gemini-2.5 and GPT-4o. Interestingly, our BISCUIT maintains strong generalization, showing minimal degradation on other downstream tasks.

Text-to-Audio (TTA) generation has made rapid progress, but current evaluation methods remain narrow, focusing mainly on perceptual quality while overlooking robustness, generalization, and ethical concerns. We present TTA-Bench, a comprehensive benchmark for evaluating TTA models across functional performance, reliability, and social responsibility. It covers seven dimensions including accuracy, robustness, fairness, and toxicity, and includes 2,999 diverse prompts generated through automated and manual methods. We introduce a unified evaluation protocol that combines objective metrics with over 118,000 human annotations from both experts and general users. Ten state-of-the-art models are benchmarked under this framework, offering detailed insights into their strengths and limitations. TTA-Bench establishes a new standard for holistic evaluation of TTA systems.

Recent advances in speech large language models (e.g., GPT-4o) have enabled end-to-end spoken interactions, yet their robustness in realworld applications remains unclear, where systems must assist users in completing specific tasks under complex conditions such as multiturn, ambiguous, and often spontaneous speech, as well as natural alternation between speech and text.Task-oriented dialogue (TOD) offers a realistic scenario to evaluate whether models can effectively help users accomplish such task-oriented goals, but existing benchmarks are mainly text-based, and the few speech datasets are limited to English and often neglect spontaneous disfluencies and speaker diversity.To address this gap, we introduce RealTalk-CN, the first Chinese multi-turn, multi-domain speech-text TOD dataset, containing 5.4k dialogues (60K turns, ~150 hours) of real human-to-human recordings with detailed annotations for dialogue states, disfluency types, and speaker characteristics.Based on this dataset, we propose a cross-modal interaction task supporting dynamic speech-text switching and a comprehensive evaluation protocol assessing robustness to disfluencies, sensitivity to speaker variation, and cross-domain generalization.Experiments on state-of-the-art models demonstrate the challenges posed by RealTalk-CN and establish its value as a benchmark for developing reliable and fair Speech LLMs in real-world deployments.The dataset and evaluation framework are available 1 to encourage further research.

Diffusion-based text-to-audio (TTA) generation has made substantial progress, leveraging latent diffusion model (LDM) to produce high-quality, diverse and instruction-relevant audios. However, beyond generation, the task of audio editing remains equally important but has received comparatively little attention. Audio editing tasks face two primary challenges: executing precise edits and preserving the unedited sections. While workflows based on LDMs have effectively addressed these challenges in the field of image processing, similar approaches have been scarcely applied to audio editing. In this paper, we introduce AudioEditor, a training-free audio editing framework built on the pretrained diffusion-based TTA model. AudioEditor incorporates Null-text Inversion and EOT-suppression methods, enabling the model to preserve original audio features while executing accurate edits. Comprehensive objective and subjective experiments validate the effectiveness of AudioEditor in delivering high-quality audio edits. Code and demo can be found at https://github.com/NKU-HLT/AudioEditor.

Incorporating visual modalities to assist Automatic Speech Recognition (ASR) tasks has led to significant improvements. However, existing Audio-Visual Speech Recognition (AVSR) datasets and methods typically rely solely on lip-reading information or speaking contextual video, neglecting the potential of combining these different valuable visual cues within the speaking context. In this paper, we release a multimodal Chinese AVSR dataset, Chinese-LiPS, comprising 100 hours of speech, video, and corresponding manual transcription, with the visual modality encompassing both lip-reading information and the presentation slides used by the speaker. Based on Chinese-LiPS, we develop a simple yet effective pipeline, LiPS-AVSR, which leverages both lip-reading and presentation slide information as visual modalities for AVSR tasks. Experiments show that lip-reading and presentation slide information improve ASR performance by approximately 8% and 25%, respectively, with a combined performance improvement of about 35%. The dataset is available at https://kiri0824.github.io/Chinese-LiPS/

Deep reinforcement learning (DRL) demonstrates its promising potential in adaptive video streaming and has recently received increasing attention. However, existing DRL-based methods for adaptive video streaming mainly use application (APP) layer information, adopt heuristic training methods, and are not robust against continuous network fluctuations. This paper aims to boost the quality of experience (QoE) of adaptive wireless video streaming by using cross-layer information, deriving a rigorous training method, and adopting effective online tuning methods with real-time data. First, we formulate a more comprehensive and accurate adaptive wireless video streaming problem as an infinite stage discounted Markov decision process (MDP) problem by additionally incorporating past and lower-layer information. This formulation allows a flexible tradeoff between QoE and computational and memory costs for solving the problem. In the offline scenario (only with pre-collected data), we propose an enhanced asynchronous advantage actor-critic (eA3C) method by jointly optimizing the parameters of parameterized policy and value function. Specifically, we build an eA3C network consisting of a policy network and a value network that can utilize cross-layer, past, and current information and jointly train the eA3C network using pre-collected samples. In the online scenario (with additional real-time data), we propose two continual learning-based online tuning methods for designing better policies for a specific user with different QoE and training time tradeoffs. The proposed online tuning methods are robust against continuous network fluctuations and more general and flexible than the existing online tuning methods. Finally, experimental results show that the proposed offline policy can improve the QoE by 6.8% to 14.4% compared to the state-of-the-arts in the offline scenario, and the proposed online policies can achieve$6.3\%$to 55.8% gains in QoE over the state-of-the-arts in the online scenario.

5G mobile communication techniques are widely applied in ubiquitous communication scenarios (e.g., smart navigation and smart transportation), which require time synchronization among mobile devices. However, the built-in time synchronization software of Android phones presents large time offsets with hundreds of milliseconds (ms), and the mainstream time synchronization techniques have specific limitations: 1) the Network Time Protocol (NTP) has too large offset to meet the real-time information interaction among mobile devices; 2) the Linux Precision Time Protocol (LinuxPTP) exists hardware dependence and cannot be implemented on Android and 5G networks; and 3) the Global Navigation Satellite System (GNSS) requires installing a hardware receiver on each mobile device. In this study, we develop a Mobile Precision Time Protocol (MobiPTP), which is hardware-independent and compatible with various network types, including Wide-Area Network (WAN), Local-Area Network (LAN), wired and wireless networks. The main challenges include signal strength instability and uplink-downlink asymmetry. We propose a dynamic time synchronization algorithm and an asymmetry compensation strategy to overcome these challenges. Regardless of high-speed mobile or crowded conditions in 5G networks, MobiPTP demonstrates an average offset of 9 ms, outperforming the NTP-based open-source software Chrony (about 30 ms). MobiPTP has been successfully deployed in multiple real-world ubiquitous communication scenarios and always demonstrates much lower offsets than Chrony.

Deep reinforcement learning (DRL) demonstrates its promising potential in the realm of adaptive video streaming. However, existing DRL-based methods for adaptive video streaming use only application (APP) layer information and adopt heuristic training methods. This paper aims to boost the quality of experience (QoE) of adaptive wireless video streaming by using lower-layer information and deriving a rigorous training method. First, we formulate a more comprehensive and accurate adaptive wireless video streaming problem as an infinite stage discounted Markov decision process (MDP) problem by additionally incorporating past and lower-layer information, allowing a flexible tradeoff between QoE and computational and memory costs for solving the problem. Then, we propose an enhanced asynchronous advantage actor-critic (eA3C) method by jointly optimizing the parameters of parameterized policy and value function. Specifically, we build an eA3C network consisting of a policy network and a value network that can utilize cross-layer, past, and current information and jointly train the eA3C network using pre-collected samples. Finally, experimental results show that the proposed eA3C method can improve the QoE by 6.8%$\sim$14.4% compared to the state-of-the-arts.

Exploring topologically distinctive trajectories provides more options for robot motion planning. Since computing time grows greatly with environment complexity, improving exploration efficiency and picking the optimal trajectory in complex environments are critical issues. To this end, this paper proposes a Graphic-and Timed-Elastic-Band-based approach (GraphicTEB) with spatial completeness and high computing efficiency. The environment is analyzed utilizing computer graphics, where obstacles are extracted as nodes and their relationships are built as edges. Three contributions are presented. 1) By assembling directed detours formed by nodes and segmented paths formed by edges, a generalized path consisting of nodes and edges derives various normal paths efficiently. 2) By multiplying two vectors starting from the obstacle point closest to the waypoint and the boundary point farthest from the waypoint, an novel obstacle gradient is introduced to guide safer optimization. 3) By assigning edges with asymmetric Gaussian model, a trajectory evaluation strategy is designed to reflect the motion tendency and motion uncertainty of dynamic obstacles. Qualitative and quantitative simulations demonstrate that the proposed GraphicTEB achieves spatial completeness, higher scene pass rate, and fastest computing efficiency. Experiments are implemented in long corridor and broad room scenarios, where the robot goes through gaps safely, finds trajectories quickly, and passes pedestrians politelyNote to Practitioners—The motivation stems from the fact that our daily cruising robot occasionally gets trapped in a corridor with piled obstacles or in a complex dynamic crowd due to the lack of a reliable trajectory. The solution is to search for more topologically distinctive trajectories and pick the optimal one. Considering that existing open-source approaches are either incomplete or highly time-consuming, a method for clustering and searching trajectories in the obstacle-occupied regions is proposed to achieve spatial completeness and high computing efficiency. In addition, an optimization technique and a trajectory selection strategy are proposed to improve motion safety. However, at present, the search is incomplete in the temporal-spatial dimension when dynamic obstacle are moving fast. How to perform a complete and fast search in temporal-spatial space will be developed in the future.

Rapid advances in single-cell chromatin accessibility sequencing (scCAS) technologies have enabled the characterization of epigenomic heterogeneity and increased the demand for automatic annotation of cell types. However, there are few computational methods tailored for cell type annotation in scCAS data and the existing methods perform poorly for differentiating and imbalanced cell types. Here, we propose CASCADE, a novel annotation method based on simulation- and denoising-based strategies. With comprehensive experiments on a number of scCAS datasets, we showed that CASCADE can effectively distinguish the patterns of different cell types and mitigate the effect of high noise levels, and thus achieve significantly better annotation performance for differentiating and imbalanced cell types. Besides, we performed model ablation experiments to show the contribution of modules in CASCADE and conducted extensive experiments to demonstrate the robustness of CASCADE to batch effect, imbalance degree, data sparsity, and number of cell types. Moreover, CASCADE significantly outperformed baseline methods for accurately annotating the cell types in newly sequenced data. We anticipate that CASCADE will greatly assist with characterizing cell heterogeneity in scCAS data analysis.

We derive approximate bit error probability (BEP) expressions for binary differential phase shift keying (BDPSK) and quadrature differential phase shift keying (QDPSK) signals in free-space optical communication system over exponentiated Weibull (EW) fading channel. By using the cumulative distribution function of EW channel and integration by parts, the conventional BEP expressions with the integral over an infinite interval are transformed to the integral over a finite interval. Based on the novel BEP expressions, we derive new, closed-form, approximate BEP expressions that involve only a gamma function, which reduce the computational complexity. The approximate BEP expressions are compared with theoretical BEP expressions under weak, moderate, and strong turbulence regimes, where the maximum gaps between the approximate and theoretical BEP expressions are 0.33 dB and 1.36 dB for BDPSK and QDPSK signals, respectively. The approximate BEP expressions are also used to analyze the loss of signal-to-noise ratio in EW fading channel.

In conventional approach, the bit error probability (BEP) expressions of$M$-ary phase-shift keying (MPSK) signals over additive white Gaussian noise (AWGN) channel are derived by averaging error probabilities for all bits of MPSK symbol. The closed-form BEP expressions of MPSK signals over AWGN channel can also be derived by using the weighted conditional symbol error probability (SEP) expressions, which has not been reported. In this paper, we derive the conditional SEP expressions of MPSK signals in terms of Gaussian$Q$and Owen's$T$functions by changing the domain of integration from a plane bounded by two rays into a quadrant. By weighting the conditional SEP expressions according to average distance spectrum, the closed-form BEP expressions of MPSK signals over AWGN channel are obtained. Only two summations involving two parameters are required. The closed-form BEP expressions of MPSK signals over Nakagami-$m$fading channel are derived by using the BEP expressions over AWGN channel and the moment generating function-based approach. The approximate BEP expressions of MPSK signals over Nakagami-$m$fading channel are obtained by using elementary functions-based approximations of Gaussian$Q$and Owen’s$T$functions. The conciseness and computational complexity of our BEP expressions are verified by the comparison with existing results.

Existing works on grant-free access, proposed to support massive machine-type communication (mMTC) for the Internet of Things (IoT), mainly concentrate on narrow band systems under flat fading. In contrast, this paper investigates massive grant-free access in a wideband system under frequency-selective fading. First, we present an orthogonal frequency division multiplexing (OFDM)-based massive grant-free access scheme. Then, we propose two different but equivalent models for the received pilot signal. Specifically, one directly models the received signal for actual devices, whereas the other can be interpreted as a signal model for virtual devices. The two signal models are insightful and essential for designing various device activity detection and channel estimation methods for OFDM-based massive grant-free access. Next, we systematically investigate statistical device activity detection under frequency-selective Rayleigh fading based on the two signal models. In particular, in the case without prior knowledge of device activities, we model device activities as deterministic but unknown binary constants and propose three maximum likelihood (ML) estimation-based device activity detection methods with different detection accuracies and computation times. In the case with prior knowledge of device activities, we model device activities as realizations of Bernoulli random variables with a known joint distribution, which appropriately incorporates the prior knowledge, and propose three maximum a posterior probability (MAP) estimation-based device activity methods, which further enhance the accuracies of the corresponding ML estimation-based methods at the cost of increased computational complexities. The proposed methods can meet diverse practical needs for OFDM-based massive grant-free access.

Storage engines for analytic databases are being developed to be setup with different devices for both cost price and performance considerations. Persistent Memory(PMem), as a novel storage device, also provides a new promising option for the hybrid storage. In this demonstration, we introduce Hockey, an efficient columnar storage engine designed for hybrid PMem-SSD storage. We first go through the design of the system and introduce how the data and meta data are structured and accessed on the PMem. The system's data placement strategy on hybrid storage is then presented. To highlight Hockey's design concerns for the hybrid storage as well as its superior performance, we develop a visual interface to demonstrate the system through three scenarios.

The impacts of channel estimation errors, inter-cell interference, phase adjustment cost, and computation cost on an intelligent reflecting surface (IRS)-assisted system are severe in practice but have been ignored for simplicity in most existing works. In this paper, we investigate a multi-antenna base station (BS) serving a single-antenna user with the help of a multi-element IRS in a multi-cell network with inter-cell interference. We consider imperfect channel state information (CSI) at the BS, i.e., imperfect CSIT, and focus on the robust optimization of the BS’s instantaneous CSI-adaptive beamforming and the IRS’s quasi-static phase shifts in two scenarios. In the scenario of coding over many slots, we formulate a robust optimization problem to maximize the user’s ergodic rate. In the scenario of coding within each slot, we formulate a robust optimization problem to maximize the user’s average goodput under the successful transmission probability constraints. The robust optimization problems are challenging two-timescale stochastic non-convex problems. In both scenarios, we obtain closed-form robust beamforming designs for any given phase shifts and more tractable stochastic non-convex approximate problems only for the phase shifts. Besides, we propose an iterative algorithm to obtain a Karush-Kuhn-Tucker (KKT) point of each of the stochastic problems for the phase shifts. It is worth noting that the proposed methods offer closed-form robust instantaneous CSI-adaptive beamforming designs which can promptly adapt to rapid CSI changes over slots and robust quasi-static phase shift designs of low computation and phase adjustment costs in the presence of imperfect CSIT and inter-cell interference. Numerical results further demonstrate the notable gains of the proposed robust joint designs over existing ones and reveal the practical values of the proposed solutions.

Device activity detection and channel estimation for grant-free massive access under frequency-selective fading have unfortunately been an outstanding problem. This paper aims to address the challenge. Specifically, we present an orthogo-nal frequency division multiplexing (OFDM)-based grant-free massive access scheme for a wideband system with one M- antenna base station (BS),$N$single-antenna Internet of Things (IoT) devices, and$P$channel taps. We obtain two different but equivalent models for the received pilot signals under frequency-selective Rayleigh fading. Based on each model, we formulate device activity detection as a non-convex maximum likelihood estimation (MLE) problem and propose an iterative algorithm to obtain a stationary point using optimal techniques. The two proposed MLE-based methods have the identical computational complexity order O(NPL2), irrespective of M, and degrade to the existing MLE-based device activity detection method when P = 1. Conventional channel estimation methods can be readily applied for channel estimation of detected active devices under frequency-selective Rayleigh fading, based on one of the derived models for the received pilot signals. Numerical results show that the two proposed methods have different preferable system parameters and complement each other to offer promising device activity detection design for grant-free massive access under frequency-selective Rayleigh fading.

The impacts of channel estimation errors, inter-cell interference, phase adjustment cost, and computation cost on an intelligent reflecting surface (IRS)-assisted system are severe in practice but have been ignored for simplicity in most existing works. In this paper, we investigate a multi-antenna base station (BS) serving a single-antenna user with the help of a multi-element IRS in the presence of channel estimation errors and inter-cell interference. We consider imperfect channel state information (CSI) at the BS, i.e., imperfect CSIT, and focus on the robust optimization of the BS's instantaneous CSI-adaptive beamforming and the IRS's quasi-static phase shifts. First, we formulate the robust optimization of the BS's instantaneous channel state information (CSI)-adaptive beamforming and IRS's quasi-static phase shifts for the ergodic rate maximization as a very challenging two-timescale stochastic non-convex problem. Then, we obtain a closed-form beamformer for any given phase shifts and a more tractable single-timescale stochastic non-convex problem only for phase shifts. Next, we propose a low-complexity stochastic algorithm to obtain quasi-static phase shifts which correspond to a KKT point of the single-timescale stochastic problem. It is worth noting that the proposed method offers a closed-form robust instantaneous CSI-adaptive beamforming design that can promptly adapt to rapid CSI changes over slots and a robust quasi-static phase shift design of low computation and phase adjustment costs in the presence of channel estimation errors and inter-cell interference. Finally, numerical results demonstrate the notable gains of the proposed robust joint design over existing ones and reveal the practical values of the proposed solutions.

In this paper, we study an intelligent reflecting surface (IRS)-assisted system where a multi-antenna base station (BS) serves a single-antenna user with the help of a multi-element IRS in the presence of interference generated by a multi-antenna BS serving its own single-antenna user. The signal and interference links via the IRS are modeled with Rician fading. To reduce phase adjustment cost, we adopt quasi-static phase shift design where the phase shifts do not change with the instantaneous channel state information (CSI). Maximum Ratio Transmission (MRT) is adopted at the two BSs to enhance the receive signals at their own users. First, we obtain a tractable expression of the ergodic rate. Then, we maximize the ergodic rate with respect to the phase shifts, corresponding to a non-convex optimization problem. We obtain a globally optimal solution under certain system parameters, and propose an iterative algorithm based on parallel coordinate descent (PCD), to obtain a stationary point under arbitrary system parameters. Finally, we numerically verify the analytical results and demonstrate the notable gains of the proposed solutions. To the best of our knowledge, this is the first work that studies the analysis and optimization of the ergodic rate of an IRS-assisted system in the presence of interference.

In this article, we study an intelligent reflecting surface (IRS)-assisted system where a multi-antenna base station (BS) serves a single-antenna user with the help of a multi-element IRS in the presence of interference generated by a multi-antenna BS serving its own single-antenna user. The signal and interference links via the IRS are modeled with Rician fading. To reduce phase adjustment cost, we adopt quasi-static phase shift design where the phase shifts do not change with the instantaneous channel state information (CSI). We investigate two cases of CSI at the BSs, namely, the instantaneous CSI case and the statistical CSI case, and apply Maximum Ratio Transmission (MRT) based on the complete CSI and the CSI of the Line-of-sight (LoS) components, respectively. Different costs on channel estimation and beamforming adjustment are incurred in the two CSI cases. First, we obtain a tractable expression of the average rate in the instantaneous CSI case and a tractable expression of the ergodic rate in the statistical CSI case. We also provide sufficient conditions for the average rate in the instantaneous CSI case to surpass the ergodic rate in the statistical CSI case, at any phase shifts. Then, we maximize the average rate and ergodic rate, both with respect to the phase shifts, leading to two non-convex optimization problems. For each problem, we obtain a globally optimal solution under certain system parameters, and propose an iterative algorithm based on parallel coordinate descent (PCD) to obtain a stationary point under arbitrary system parameters. Next, in each CSI case, we provide sufficient conditions under which the optimal quasi-static phase shift design is beneficial, compared to the system without IRS. Finally, we numerically verify the analytical results and demonstrate notable gains of the proposal solutions over existing ones. To the best of our knowledge, this is the first work that considers optimal quasi-static phase shift design for an IRS-assisted system in the presence of interference.