This paper investigates signaling by a base station equipped with a dynamic metasurface antenna (DMA) to support the transmission of multi-stream information to multiple near-field users. We consider the joint design of the low-resolution DMA elements’ frequency responses and the baseband precoder to ensure the quality-of-service (QoS) for all users in terms of their rates. First, we develop convex quadratic solver-based iterations of cubic complexity to address the computationally challenging max-min rate optimization problem involving nonsmooth large-scale mixed discrete-continuous optimization. We then opt for the soft max-min rate optimization problem, which involves smooth mixed discrete-continuous optimization, and develop closed-form expression-based iterations of scalable complexity for its computation. The latter approach is not only computationally efficient but also achieves both a high minimum user rate and sum-rate, thereby guaranteeing QoS and high overall network throughput.

This work explores signal transmission for serving multiple communication users (CUs) while simultaneously estimating multiple targets. In this context, the multi-stream signals intended for downlink CUs equipped with multiple antennas are also employed as probing signals for target estimation. We consider precoding design to ensure quality of service, quantified by the CUs’ individual rates and the mean squared error in target estimation. We develop path-following computational procedures that generate a sequence of improved feasible points by iterating closed-form expressions, ensuring convergence. As a byproduct, a computational solution for estimating the targets’ response vectors or their reflection coefficients and angles manifests, addressing long-standing open problems in estimation theory. Computational experiments not only demonstrate their consistency but also reveal that the obtained precoders produce highly directionally selective beampatterns toward the targets, even though this is not a primary objective.

This work investigates the integration of multi-target inference and multiuser communication in an active reconfigurable intelligent surface (aRIS)-assisted network. To infer the targets’ elevation and azimuth pairs and reflection coefficients from signals transmitted by a base station and reflected by the aRIS, which form computationally intractable nonlinear models, we develop a constructive minimum mean square error (MMSE) estimator based on their probability distribution functions. The resulting MSE is expressed analytically as a deterministic function of the probing signal, enabling its optimization. We then formulate the problem of jointly designing a beamformer and the aRISs power-amplified reconfigurable elements to ensure both accurate target inference and fair user rates. A computational program using closed-form updates is developed. Numerical results demonstrate a flexible trade-off between inference accuracy and achieved user rates.

Despite the impressive chain-of-thought (CoT) reasoning ability of large language models (LLMs), its underlying mechanisms remains unclear.In this paper, we explore the inner workings of LLM's CoT ability via the lens of neurons in the feed-forward layers.We propose an efficient method to identify reasoningcritical neurons by analyzing their activation patterns under reasoning chains of varying quality.Based on it, we devise a rather simple intervention method that directly stimulates these reasoning-critical neurons, to guide the generation of high-quality reasoning chains.Extended experiments validate the effectiveness of our method and demonstrate the critical role these identified neurons play in CoT reasoning.

Large Language Models fine-tuned with Reinforcement Learning from Human Feedback (RLHF-LLMs) can over-rely on aligned preferences without truly gaining self-knowledge, leading to hallucination and biases. If an LLM can better access its knowledge and know what it knows, it can avoid making false or unsupported claims. Therefore, it is crucial to evaluate whether LLMs have the ability to know what they know, as it can help to ensure accuracy and faithfulness in real-world applications. Inspired by research in Educational Psychology, surface learners who don’t really know are easily affected by teacher and peer guidance, we treat LLM as a student, incorporate role guidance in prompts to explore whether LLMs really know. Specifically, we propose a novel strategy called Role-Guided and Self-Reflection (RoSe) to fully assess whether LLM “knows it knows”. We introduce multiple combinations of different roles and strong reminder in prompts combined with self-reflection to explore what local information in prompt LLMs rely on and whether LLMs remain unaffected by external guidance with varying roles. Our findings reveal that LLMs are very sensitive to the strong reminder information. Role guidance can help LLMs reduce their reliance on strong reminder. Meanwhile, LLMs tend to trust the role of authority more when guided by different roles. Following these findings, we propose a double-calibrated strategy with verbalized confidence to extract well-calibrated data from closed-source LLM and fine-tune open-source LLMs. Extensive experiments conducted on fine-tuning open-source LLMs demonstrate the effectiveness of double-calibrated strategy in mitigating the reliance of LLMs on local information. For a thorough comparison, we not only employ public JEC-QA and openBookQA datasets, but also construct EG-QA which contains English Grammar multiple-choice question-answering and 14 key knowledge points for assessing self-knowledge and logical reasoning.

In this paper, we investigate a full-duplex (FD) UAV-assisted multi-user system under multiple malicious jammers and propose a robust online scheme for secure communication with mobile downlink and uplink users. A random mobility model is adopted to simulate user movement, and the problem is formulated as a two-stage online optimization framework comprising a present-point and a prediction-point problem. To tackle the non-convexity, we develop inner-approximation algorithms using successive convex approximation (SCA) and the S-procedure. Simulation results verify the effectiveness of the proposed method.

Mathematical reasoning is an important capability of large language models~(LLMs) for real-world applications. To enhance this capability, existing work either collects large-scale math-related texts for pre-training, or relies on stronger LLMs (\eg GPT-4) to synthesize massive math problems. Both types of work generally lead to large costs in training or synthesis. To reduce the cost, based on open-source available texts, we propose an efficient way that trains a small LLM for math problem synthesis, to efficiently generate sufficient high-quality pre-training data. To achieve it, we create a dataset using GPT-4 to distill its data synthesis capability into the small LLM. Concretely, we craft a set of prompts based on human education stages to guide GPT-4, to synthesize problems covering diverse math knowledge and difficulty levels. Besides, we adopt the gradient-based influence estimation method to select the most valuable math-related texts. The both are fed into GPT-4 for creating the knowledge distillation dataset to train the small LLM. We leverage it to synthesize 6 million math problems for pre-training our JiuZhang3.0 model. The whole process only needs to invoke GPT-4 API 9.3k times and use 4.6B data for training. Experimental results have shown that JiuZhang3.0 achieves state-of-the-art performance on several mathematical reasoning datasets, under both natural language reasoning and tool manipulation settings. Our code and data will be publicly released in \url{https://github.com/RUCAIBox/JiuZhang3.0}.

This paper considers a scenario involving a network where two reconfigurable intelligent surfaces (RISs) contribute to enhancing the security of multi-user secure downlink communication, even in the presence of multiple potential eavesdroppers. The objective is to optimize the base station (BS)’s beamforming and the quantized programmable reflecting elements (PREs) of both RISs to maximize the geometric mean of secrecy rates (GM-SECR). To tackle this intricate non-convex penalized optimization problem, the paper introduces an alternating descent iteration algorithm based on closed-form solutions. Through simulations, the study highlights the advantages presented by the proposed double-RIS system and validates the efficacy of the algorithm. Notably, the results demonstrate a marked enhancement in achieving fair distributions of secrecy rates.

This paper investigates the unmanned aerial vehicle (UAV)-assisted communication network with multiple downlink users (DLUs) and uplink users (ULUs) in the presence of multiple malicious jammers. To guarantee fairness among the users and their uplink and downlink communication throughput, we aim to maximize the minimum average throughput by jointly optimizing the scheduling of ULUs/DLUs, three dimensional (3D) trajectory and the UAV transmission power. Although the optimization problem is computationally intractable due to its non-convexity, we develop an iterative algorithm based on the block coordinate descend approach and the successive convex approximation technique to solve the problem efficiently. Numerical outcomes show that our proposed algorithm can improve throughput significantly over several benchmark schemes.

This paper is concerned with distributed consensus control of unknown multiagent systems. As the system’s dynamics is unknown, an adaptive Fourier decomposition (AFD) based iterative learning control (ILC) dynamics adaptive matching method in frequency domain is put forward to deal with it. First, large amounts of input and output measurement data are used to estimate the frequency domain characteristics of the system by Takenaka-Malmquist functions. Second, convert the traditional time domain ILC to the frequency domain to establish a matching relationship with the estimated frequency domain features. Then, an adaptive iterative learning rate is constructed to achieve the optimal convergence at each sampling point. The feasibility of the proposed algorithm is guaranteed by the convergence of AFD in Hardy space$H^{2}(\mathbb {D})$under the maximum selection principle. Compared with the reinforcement learning data-driven control scheme, the method proposed in this paper has obvious advantages in the control accuracy and convergence efficiency. In addition, this paper takes two kinds of denoising algorithms based on unwinding AFD to deal with the multi-agent systems with channel noise. Finally, the feasibility and effectiveness of the developed method are verified by a series of simulations.

In this paper, we present several theoretic results on density evolution (DE) for nonbinary spatially-coupled low-density parity-check (SC-LDPC) ensembles when the transmission takes place on the binary erasure channel (BEC). In specific, we establish the duality rule for entropy for the nonbinary variable-node (VN) and check-node (CN) operators in such a scenario. We define the partial order between densities and show that the VN and CN operators exhibit the property of partial order preservation. More importantly, we explicitly construct the potential functions for uncoupled and coupled DE recursions, the forms of which almost coincide with those for binary LDPC and SC-LDPC ensembles over general binary memoryless symmetric (BMS) channels. These theoretic findings greatly facilitate the proof of threshold saturation for nonbinary SC-LDPC ensembles on the BEC. Finally, we develop the threshold saturation theorem and its converse, following the lines established by S. Kumar et al.

In this paper, an unmanned aerial vehicle (UAV) enabled secure downlink communication is considered, where a single-antenna UAV serves multiple ground users facilitated by orthogonal frequency-division multiple access (OFDMA), in the presence of an eavesdropper (EV) with imperfect channel state information. To enhance the secrecy rate (SR), we employ a power splitting approach, where part of the transmit power is used for communication while the rest is used for jamming. We maximize the average secrecy rate (ASR) by jointly optimizing the bandwidth, trajectory, power allocation, and power splitting ratio. To tackle this non-convex and computationally intractable optimization problem, we propose a novel algorithm by employing successive convex approximation, block coordinate descend and$S$-procedure. Numerical results show that our proposed joint optimization algorithm outperforms the benchmark schemes.

Unmanned aerial vehicle (UAV)-aided two-way relaying networks are designed, where a UAV is deployed to assist multiple pairs of users in their information exchange. There are two basic approaches for the user pairs' information exchange within a single time slot via the UAV relay. The first approach is based on full-duplex, where all participants operate in the full-duplex mode to transmit and receive signals simultaneously. However, all transceivers have to operate in the face of severe self-interference, which cannot be completely suppressed. The second approach is based on conventional half-duplex, where the users send their information to the UAV within a certain fraction of the time slot, and the UAV relays them within the remaining fraction to avoid the self-interference. In either approach, the joint bandwidth and power allocation maximizing the sum information exchange throughput under realistic resource and user throughput constraints poses a complex nonconvex problem. New inner approximations are proposed for developing path-following algorithms for their computation. Our numerical results show that the time-fraction-based half-duplex approach clearly outperforms the high-complexity full-duplex approach.

Under both long (infinite) and short (finite) blocklength transmissions, this paper considers physical layer security for a wireless interference network of multiple transmitter-user pairs, which is overheard by multiple eavesdroppers (EVs). The EVs are assumed to have better channel conditions than the legitimate users (UEs), making the conventional transmission unsecured. The paper develops a novel time-fraction based transmission, under which the information is transmitted to the UEs within a fraction of the time slot and artificial noise (AN) is transmitted within the remaining fraction to counter the strong EVs' channels. Based on channel distribution information of UEs and EVs, the joint design of transmit beamforming, time fractions and AN power allocation to maximize the worst users' secrecy rate is formulated in terms of nonconvex problems. Path-following algorithms of low complexity and rapid convergence are proposed for their solution. Simulations are provided to demonstrate the viability of the proposed methodology.

This paper considers a wireless interference network in which the communication between multiple transmitter-user pairs is overheard by multiple eavesdroppers (EVs). Based on knowledge of the channel distribution, the goal is to maximize the worst users' secrecy rate under both long (infinite) blocklength and short (finite) blocklength transmissions. Under long blocklength transmission, the performance of the existing algorithms is unsatisfactory when the wiretapped channels are sufficiently strong. To address this drawback, we adopt a time-fraction based information and artificial noise (AN) transmission, under which first the information is transmitted within the initial fraction of the time slot and then AN is transmitted within the remaining fraction. Accordingly, the problem of join optimization of the time fractions, transmit power, and AN power to maximize the minimum secrecy rate is proposed and computed by a path-following algorithm, which iterates feasible points and converges at least to a locally optimal solution. A similar problem under short blocklength transmission is also proposed and computed. The provided simulations results clearly show the merits of the proposed approach.

This paper proposes Han-Kobayashi signaling (HKS), under which each pair of users decodes a common message to improve their throughput, for UAV-enabled multi-user communication. Given that only a single transmit antenna is used and thus there is no null space of users' channels for inserting an artificial noise that would effectively help to jam an eavesdropper without interfering the users' desired signals, a new information and artificial noise transfer scheme to address physical layer security (PLS) for the considered networks is investigated. Under this scheme, the UAV sends the confidential information to its users within a fraction of the time slot and sends the artificial noise within the remaining fraction. Accordingly, the problem of jointly optimizing the time-fraction, bandwidth and power allocation to maximize the users' worst secrecy throughput is formulated. New inner approximations are proposed for developing path-following algorithms for its computation. Simulation shows that the proposed information and artificial noise transfer enables not only HKS but also orthogonal multi-access and nonorthogonal multi-access to provide PLS for UAV-enabled communication even when the eavesdropper is in the best channel condition. HKS outperforms the other two schemes in terms of users' worst secrecy throughput.

Based on the knowledge of the channel distributions of a multi-input single-output wireless network of multiple transmitter-user pairs overheard by an eavesdropper, this letter develops an outage-aware beamforming design to optimize the users' quality-of-service (QoS) in terms of their secrecy rates. This is a very computationally difficult problem with a nonconcave objective function and nonlinear equality constraints in beamforming vectors. A path-following algorithm of low-complexity and rapid convergence is proposed for computation, which is also extended to solving the problem of maximizing the network's secure energy efficiency under users' QoS constraints. Numerical examples are provided to verify the efficiency of the proposed algorithms.

Considering a multi-user interference network with an eavesdropper, this paper aims at the power allocation to optimize the worst secrecy throughput among the network links or the secure energy efficiency in terms of achieved secrecy throughput per Joule under link security requirements. Three scenarios for the access of channel state information are considered: the perfect channel state information; partial channel state information with channels from the transmitters to the eavesdropper exponentially distributed; and not perfectly known channels between the transmitters and the users with exponentially distributed errors. The paper develops various path-following procedures of low complexity and rapid convergence for the optimal power allocation. Their effectiveness and viability are illustrated through numerical examples. The power allocation schemes are shown to achieve both high secrecy throughput and energy efficiency.

This paper considers the joint design of user power allocation and relay beamforming in relaying communications, in which multiple pairs of single-antenna users exchange information with each other via multiple-antenna relays in two time slots. All users transmit their signals to the relays in the first time slot while the relays broadcast the beamformed signals to all users in the second time slot. The aim is to maximize the system's energy efficiency (EE) subject to quality-of-service (QoS) constraints in terms of exchange throughput requirements. The QoS constraints are nonconvex with many nonlinear cross-terms, so finding a feasible point is already computationally challenging. The sum throughput appears in the numerator while the total consumption power appears in the denominator of the EE objective function. The former is a nonconcave function and the latter is a nonconvex function, making fractional programming useless for EE optimization. Nevertheless, efficient iterations of low complexity to obtain its optimized solutions are developed. The performance of the multiple-user and multiple-relay networks under various scenarios is evaluated to show the merit of the proposed method.

This paper considers power allocation between data and pilot symbols for an orthogonal frequency division multiplexing (OFDM) system in a high-speed train (HST) environment. The channel gains are very quickly time-varying within an OFDM symbol so both their estimation and symbol detection must be simultaneously implemented with unavoidable inter-carrier interferences. The average channel complex gains are estimated and are used to calculate the basis expansion model (BEM) coefficients of the HST channel in data detection. We choose the effective signal-to-interference and noise ratio (SINR) in symbol detection as the cost function and propose the effective algorithm to maximize it. The simulation results confirm the viability of our proposed algorithm.