In this paper, we introduce a novel approach for addressing the multi-objective optimization problem in large language model merging via black-box multi-objective optimization algorithms. The goal of model merging is to combine multiple models, each excelling in different tasks, into a single model that outperforms any of the individual source models. However, the effectiveness of conventional model merging methods is constrained by human intuition or domain knowledge. While existing optimization-based model merging methods can automatically search for model merging parameter configurations, they often struggle to find a satisfactory configuration within a limited evaluation budget. To address this challenge, we propose a novel and sample-efficient automated model merging method, named MM-MO. This method leverages multi-objective Bayesian optimization algorithms to autonomously search for great merging configurations across various tasks. In MMMO, we proposed an enhanced acquisition strategy and an auxiliary optimization objective to improve the search process. Our enhanced acquisition strategy integrates a weak-to-strong method to refine the acquisition function, enabling previously evaluated superior configurations to guide the search for new ones. Meanwhile, Fisher information is utilized to further filter these configurations, increasing the possibility of finding high-quality merging configurations. Additionally, we design a sparsity metric as an auxiliary optimization objective, further enhance the models generalization performance across different tasks. We conducted comprehensive experiments with other mainstream model merging methods, demonstrating that the proposed MMMO algorithm is competitive and effective in achieving high-quality model merging.

Multi-objective Bayesian optimization (MOBO) has shown promising performance on various expensive multi-objective optimization problems (EMOPs). However, effectively modeling complex distributions of the Pareto optimal solutions is difficult with limited function evaluations. Existing Pareto set learning algorithms may exhibit considerable instability in such expensive scenarios, leading to significant deviations between the obtained solution set and the Pareto set (PS). In this paper, we propose a novel Composite Diffusion Model based Pareto Set Learning algorithm (CDM-PSL) for expensive MOBO. CDM-PSL includes both unconditional and conditional diffusion model for generating high-quality samples efficiently. Besides, we introduce a weighting method based on information entropy to balance different objectives. This method is integrated with a guiding strategy to appropriately balancing different objectives during the optimization process. Experimental results on both synthetic and real-world problems demonstrates that CDM-PSL attains superior performance compared with state-of-the-art MOBO algorithms.

Empathetic response generation necessitates the integration of emotional and intentional dynamics to foster meaningful interactions.Existing research either neglects the intricate interplay between emotion and intent, leading to suboptimal controllability of empathy, or resorts to large language models (LLMs), which incur significant computational overhead.In this paper, we introduce ReflectDiffu, a lightweight and comprehensive framework for empathetic response generation.This framework incorporates emotion contagion to augment emotional expressiveness and employs an emotion-reasoning mask to pinpoint critical emotional elements.Additionally, it integrates intent mimicry within reinforcement learning for refinement during diffusion.By harnessing an intent twice reflect mechanism of Exploring-Sampling-Correcting, ReflectDiffu adeptly translates emotional decision-making into precise intent actions, thereby addressing empathetic response misalignments stemming from emotional misrecognition.Through reflection, the framework maps emotional states to intents, markedly enhancing both response empathy and flexibility.Comprehensive experiments reveal that ReflectDiffu outperforms existing models regarding relevance, controllability, and informativeness, achieving stateof-the-art results in both automatic and human evaluations.

Large language models (LLMs) have shown remarkable performance in reasoning tasks but face limitations in mathematical and complex logical reasoning.Existing methods to improve LLMs' logical capabilities either involve traceable or verifiable logical sequences that generate more reliable responses by constructing logical structures yet increase computational costs, or introduces rigid logic template rules, reducing flexibility.In this paper, we propose Reversal of Thought (RoT), a plug-and-play and cost-effective reasoning framework designed to enhance the logical reasoning abilities of LLMs during the warm-up phase prior to batch inference.RoT utilizes a Preference-Guided Reverse Reasoning warm-up strategy, which integrates logical symbols for pseudocode planning through meta-cognitive mechanisms and pairwise preference self-evaluation to generate task-specific prompts solely through demonstrations, aligning with LLMs' cognitive preferences shaped by RLHF.Through reverse reasoning, we utilize a Cognitive Preference Manager to assess knowledge boundaries and further expand LLMs' reasoning capabilities by aggregating solution logic for known tasks and stylistic templates for unknown tasks.Experiments across various tasks demonstrate that RoT surpasses existing baselines in both reasoning accuracy and efficiency.

Artificial Intelligence (AI) technology has been increasingly integrated into Radio Access Networks (RAN), thereby boosting the intelligence capabilities of the air interface in mobile communications. Specifically, AI-driven beam management technology achieves multiple performance breakthroughs. These include spectral efficiency improvement and coverage quality enhancement, enabled by dynamic optimization of beam direction and intelligent resource multiplexing. This paper proposes a differentiated service datadriven intelligent adjustment method for 5G-A broadcast beams. By considering service-level performance requirements and traffic popularity, the method dynamically and intelligently adjusts broadcast beam parameters. The method aims to provide differential network coverage guarantee for different services under fixed resources and enhance the precision of broadcast beam adjustment.

During the During the National People's Congress and the Chinese Political Consultative Conference in 2024, the ‘low-altitude economy’ was included in the government work report as a significant factor driving new quality productive forces. In the New Radio (NR) network, when a terminal is accessed, the base station uses SSB (Synchronization Signal Block) beam sweeping to detect the optimal beam for the terminal. After the terminal accesses and obtains the configuration information of the reference signal, it feeds back the channel state information (CSI), and the base station uses the optimal beam from CSI-RS (Channel State Information Reference Signal) beam sweeping. In low-altitude communications, 5G antennas flexibly configure the number of beams, considering horizontal and vertical dimensions. Combining SSB frequency offset technology, the SSB frequency points of the low-altitude network can be staggered with the configuration of the ground network, forming a virtual airground heterogeneous frequency network. This approach enhances performance by reducing handover times and interference.

With the rapid development of communication technologies, the need for accurate and efficient complaint auxiliary analysis (CAA) among mobile network optimization personnel is growing. However, most existing research solutions focus only on text data and structured data, with few incorporating user complaint speech data. To address this, the authors propose a CAA scheme for mobile network based on multi-modal generative Large Language Model (LLM). By integrating speech, text, and other multi-modal data, the proposed scheme aims to accurately understand and efficiently analyze user complaint information. The proposed approach consists of three key components. First, the Whisper model is employed to transform user complaint speech data into structured textual representations. Subsequently, a self-constructed domain-specific dictionary is integrated with an Attention-based mechanism to train a Key Information Extraction (KIE) model, thereby enhancing its semantic comprehension performance. Finally, keyword-based knowledge derived from knowledge graphs is combined with expert-defined rules to train a Generative Programs and Recommendations (GPR) model, enabling the system to deliver more accurate and professional Customer Assistance Automation (CAA) solutions. Experimental evaluations demonstrate that the proposed framework exhibits superior generalization capabilities.

This paper investigates the workflow and requirement of telecommunications operators in handling 4G/5G user network quality complaints and proposes a solution that combines large and small models to achieve more intelligent complaint handling. The large model is responsible for comprehensively analyzing unstructured data such as user complaint texts, extracting key information, and understanding user intentions. Small models are used for indepth processing of structured data related to network performance indicators, conducting root cause analysis, and providing targeted solutions. The models and systems are applied to current network operations, significantly reducing network maintenance optimization work orders, saving labor costs, and improving work efficiency.

This paper analyzes the typical service requirements of low-altitude scenarios and proposes an intelligent weight optimization scheme for 5G beams in low-altitude scenarios using artificial bee colonies and genetic algorithms. Based on key indicators such as coverage quality, interference level, and service perception, joint optimization was conducted and validated in low-altitude networking pilot areas, resulting in significant improvements in computational efficiency and optimization results. This scheme achieved the optimal solution for the weight of contiguous areas, resulting in sound application effects.

In mobile communication networks, host classification plays a critical role in constructing user profiles and ensuring network security. Traditional approaches, which rely on rule-based matching and shallow feature engineering, face significant limitations in coping with the high-frequency dynamic variations of hostnames and the labor-intensive maintenance of manual rules. To address these challenges, this paper proposes a novel frequency-aware hybrid-granularity tokenization method, specifically designed to capture both the semantic structure and statistical patterns of hostnames. By leveraging semi-supervised learning on large-scale host sequence data collected from real-world network environments, the proposed method enables effective service classification through vectorized host representations. This work not only offers an efficient and scalable solution for host analysis in personalized recommendation systems and mobile network security but also provides valuable insights into the design of pretrained tokenizers tailored for dynamic data scenarios.

Temporal action proposal generation plays a vital role in the analysis of untrimmed videos and has garnered growing interest from researchers. Nevertheless, the presence of long-term temporal dependencies and the large variation in action durations within untrimmed videos pose significant challenges for accurately localizing action boundaries. To overcome the aforementioned issues, we design a novel Transformer-based Temporal Feature Pyramid Network (TTFPN) tailored for generating action proposals. Specifically, we introduce a local transformer to capture longterm temporal information while reducing computational complexity through the substitution of conventional selfattention with a localized variant. Subsequently, a temporal feature pyramid is built to produce multi-scale representations, enabling the model to effectively handle action instances of varying durations. Based on this temporal feature pyramid, we employ a convolutional network-based predictor to generate action proposals in an anchor-free manner. We evaluate TTFPN on THUMOS14, a standard benchmark for temporal action detection, to validate its effectiveness. The results show that TTFPN achieves competitive performance and significantly outperforms previous methods.

With the gradual commercialization of 5G VoNR, VoLTE and VoNR will become the main methods of voice services. How to efficiently evaluate the quality of voice service is the focus of telecom operators. This paper proposes an intelligent combined evaluation method of VoLTE and VoNR voice quality based on artificial neural network. In the proposed method, the artificial neural network model is fitted by the call level time slice sample data of voice, and then the prediction model is established. The prediction results of voice quality of mobile networks are obtained by using the prediction model at call level, grid level and area level. Meanwhile, the proposed method can address the shortcomings of traditional evaluation method based on road test, such as high cost, low timeliness and limited area. Finally, through theoretical verification and comparison with the real test results, the effectiveness of the prediction method is verified.

Recently, location-based service has become a hot research topic. Mobile communication data records abundant information about users’ temporal and spatial characteristics. By modeling the users’ mobility based on mobile communication data, this can assist to understand human user patterns more accurately and deeply. Initially, this paper introduces three mainstream algorithms for user mobility modeling. Then this paper proposes a novel Markov chain based user mobility prediction scheme. The proposed scheme is implemented through four stages, including time and space division, Markov property examination, transition probability matrix calculation, Markov model weighting. Experimental results show that the proposed scheme can achieve higher accuracy compared with the traditional algorithms.

The 6G communication system is expected to achieve seamless global coverage. The orthogonal time-frequency space (OTFS) is generally considered as the main candidate waveform for 6G communication. Especially, in the air-space-ground integrated communication system, OTFS is more suitable for air interface modulation waveforms for high mobility communication scenarios than OFDM system. This paper focuses on OTFS technology, which conveniently adapts to the channel constantly changing via modulating information. In the paper, comparative analysis under different rate scenarios is performed, and potential future application scenarios are proposed, such as applying artificial intelligence based on vehicle network.

MDT data have been widely used for 4G/5G wireless network coverage estimation. Whereas the sparsity of the MDT data makes coverage rate bias when it applied into realistic network coverage analysis. To achieve a more precise coverage estimation, this paper proposes an approach that adding geographical and landform information to network coverage estimation in order to refine the coverage rate. An attention U-Net model was applied to landforms recognition from online satellite map with low cost. It can effectively assists telecom operators to filter out areas, which are users inaccessible or do not require signal coverage.

This paper investigates on the remote interference problem in the TDD network and proposes an intelligent 5G Remote Interference Avoidance and Clustering Scheme (RIAC), on the basis of RIM-RS (remote interference management-reference signal) and clustering algorithm. This paper adopts the GBLA-DBSACN (the grid-based local adaptive DBSCAN) algorithm based on the traditional DBSCAN algorithm (Density—Based Spatial Clustering of Application with Noise) to improve the accuracy of interference base station (BS) clustering, which considers the dispersion of interference sources. This scheme helps to locate interference problems and potential sources through testing in the existing network quickly and effectively. By taking corresponding optimization means for these problems, network operators can effectively reduce the interference level in the target area and improve the quality of network construction.

The high popularity of 5G has spawned a large number of emerging application scenarios and diversified business models, meanwhile, it also leads to the increase in network capacity. The research on 5G network capacity has become an important topic to improve the user perception. This paper analyzes the future capacity trend and development characteristic model of 5G, and then determines the four dimensions for evaluating 5G network capacity. Based on each dimension, this paper locates key indicators, and creatively puts forward the concept of experience satisfaction. Furthermore, this paper researches and recommends the capacity expansion thresholds for 3.5G and 2.1G respectively, using the big data fitting method. In addition, this paper also finds the internal relationship between these key indicators, and give the recommended capacity expansion threshold for each type of cell. A reasonable and accurate capacity expansion threshold is can effectively use the limited capacity expansion investment as well as improve user perception of 5G network.