Fine-tuning is the process of adapting the pre-trained large language models (LLMs) for downstream tasks. Due to substantial model parameters, fine-tuning LLM on mobile devices demands considerable memory resources, and suffers from high communication overhead and fine-tuning delay. In this paper, we propose an efficient LLM fine-tuning scheme in wireless networks, named <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">S</u>plit <underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F</u>ine-<underline xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">T</u>uning (SFT), which can accommodate LLM fine-tuning on mobile devices. Specifically, an LLM is split into a server-side part on the edge server and a device-side part on the mobile device to satisfy the device-side memory constraint. All devices share a server-side model and perform parallel fine-tuning to reduce fine-tuning delay. In addition, to reduce communication overhead incurred by data exchange between devices and the edge server, we propose an activation data compression scheme by jointly leveraging sparsification, stochastic quantization, and lossless encoding methods. Furthermore, we formulate a fine-tuning delay minimization problem under model accuracy and device-side memory constraints, taking device heterogeneity and channel dynamics into account. To solve the problem, the nonlinear mixed-integer problem is decoupled into two subproblems in different timescales. A two-timescale resource management algorithm is proposed to jointly optimize the compression rate and transformer block allocation in the large timescale using the augmented Lagrangian method, and determine spectrum resource allocation in the small timescale via sequential quadratic programming. Extensive simulation results demonstrate that the proposed scheme can reduce the fine-tuning delay by up to 66.4% and communication overhead by 93.6% compared to state-of-the-art benchmarks, while satisfying device-side memory and model accuracy constraints.

The outstanding performance of Large Multimodal Models (LMMs) has made them widely applied in vision-related tasks. However, various corruptions in the real world mean that images will not be as ideal as in simulations, presenting significant challenges for the practical application of LMMs. To address this issue, we introduce R-Bench, a benchmark focused on the Real-world Robustness of LMMs. Specifically, we: (a) model the complete link from user capture to LMMs reception, comprising 33 corruption dimensions, including 7 steps according to the corruption sequence, and 7 groups based on low-level attributes; (b) collect reference/distorted image dataset before/after corruption, including 2,970 question-answer pairs with human labeling; (c) propose comprehensive evaluation for absolute/relative robustness and benchmark 20 mainstream LMMs. Results show that while LMMs can correctly handle the original reference images, their performance is not stable when faced with distorted images, and there is a significant gap in robustness compared to the human visual system. We hope that R-Bench will inspire improving the robustness of LMMs, extending them from experimental simulations to the real-world application.

Speech enhancement plays an essential role in various applications, and the integration of visual information has been demonstrated to bring substantial advantages. However, the majority of current research concentrates on the examination of facial and lip movements, which can be compromised or entirely inaccessible in scenarios where occlusions occur or when the camera view is distant. Whereas contextual visual cues from the surrounding environment have been overlooked: for example, when we see a dog bark, our brain has the innate ability to discern and filter out the barking noise. To this end, in this paper, we introduce a novel task, i.e. Scene-aware Audio-Visual Speech Enhancement (SAV-SE. To our best knowledge, this is the first proposal to use rich contextual information from synchronized video as auxiliary cues to indicate the type of noise, which eventually improves the speech enhancement performance. Specifically, we propose the VC-S <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{2}$</tex-math></inline-formula> E method, which incorporates the Conformer and Mamba modules for their complementary strengths. Extensive experiments are conducted on public MUSIC, AVSpeech and AudioSet datasets, where the results demonstrate the superiority of VC-S <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$^{2}$</tex-math></inline-formula> E over other competitive methods. We will make the source code publicly available. Project demo page: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://AVSEPage.github.io/</uri>

Being aware of the channel and its properties is critical for coherent transmission in massive multiple-input multiple-output (M-MIMO) systems due to the large channel dimension in the space domain. In cell-free (CF) systems, the channel dimension increases further as each user is served by multiple access points, with a significant burden on signal processing. Angle domain transmission and channel maps promise to alleviate this burden by reducing channel dimensions in the angle domain and providing a priori channel information through channel measurements and modeling, respectively. In this paper, we propose a channel map-based angle domain multiple access scheme for the uplink CF M-MIMO communications. First, we propose an angle domain data reception scheme constituting receive combining and large-scale fading decoding to maximize spectral efficiency. Then, we derive an initial access criterion utilizing the angle domain channel similarity between users, based on which we propose pilot assignment and access point selection schemes for better trade-offs between spectral and energy efficiency. Finally, we construct two channel map-based transmission mechanisms by wielding different levels of channel information, where a tailored data reception scheme with a newly derived spectral efficiency upper bound is also proposed for quantitative evaluation. Simulation results show that the proposed channel map-based angle domain schemes outperform their space domain alternatives and the schemes without using channel maps regarding spectral and energy efficiency.

Cell-free massive multi-input multi-output (CF-mMIMO) systems have emerged as a promising paradigm for next-generation wireless communications, offering enhanced spectral efficiency and coverage through distributed antenna arrays. However, the non-linearity of power amplifiers (PAs) in these arrays introduce spatial distortion, which may significantly degrade system performance. This paper presents the first investigation of distortion-aware beamforming in a distributed framework tailored for CF-mMIMO systems, enabling pre-compensation for beam dispersion caused by nonlinear PA distortion. Using a third-order memoryless polynomial distortion model, the impact of the nonlinear PA on the performance of CF-mMIMO systems is firstly analyzed by evaluating the signal-to-interference-noise-and-distortion ratio (SINDR) at user equipment (UE). Then, we develop two distributed distortion-aware beamforming designs based on ring topology and star topology, respectively. In particular, the ring-topology-based fully-distributed approach reduces interconnection costs and computational complexity, while the star-topology-based partially-distributed scheme leverages the superior computation capability of the central processor to achieve improved sum-rate performance. Extensive simulations demonstrate the effectiveness of the proposed distortion-aware beamforming designs in mitigating the effect of nonlinear PA distortion, while also reducing computational complexity and backhaul information exchange in CF-mMIMO systems.

Text-to-Speech (TTS) has recently seen great progress in synthesizing high-quality speech owing to the rapid development of parallel TTS systems. Yet producing speech with naturalistic prosodic variations, speaking styles, and emotional tones remains challenging. In addition, many existing parallel TTS models often struggle with identifying optimal monotonic alignments since speech and duration generation typically occur independently. Here, we propose StyleTTS, a style-based generative model for parallel TTS that can synthesize diverse speech with natural prosody from a reference speech utterance. Using our novel Transferable Monotonic Aligner (TMA) and duration-invariant data augmentation, StyleTTS significantly outperforms other baseline models on both single and multi-speaker datasets in subjective tests of speech naturalness and synthesized speaker similarity. It also demonstrates higher robustness and emotional similarity to the reference speech as indicated by word error rate (WER) and acoustic feature correlations. Through self-supervised learning, StyleTTS can generate speech with the same emotional and prosodic tone as the reference speech without needing explicit labels for these categories. In addition, when trained with a large number of speakers, our model can perform zero-shot speaker adaption. The source code and audio samples can be found on our demo page at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://styletts.github.io/</uri>.

Recent research in speaker verification has increasingly focused on achieving robust and reliable recognition under challenging channel conditions and noisy environments. Identifying speakers in radio communications is particularly difficult due to inherent limitations such as constrained bandwidth and pervasive noise interference. To address this issue, we present a Channel Robust Speaker Learning (CRSL) framework that enhances the robustness of the current speaker verification pipeline, considering data source, data augmentation, and the efficiency of model transfer processes. Our framework introduces an augmentation module that mitigates bandwidth variations in radio speech datasets by manipulating the bandwidth of training inputs. It also addresses unknown noise by introducing noise within the manifold space. Additionally, we propose an efficient fine-tuning method that reduces the need for extensive additional training time and large amounts of data. Moreover, we develop a toolkit for assembling a large-scale radio speech corpus and establish a benchmark specifically tailored for radio scenario speaker verification studies. Experimental results demonstrate that our proposed methodology effectively enhances performance and mitigates degradation caused by radio transmission in speaker verification tasks.