Integrated circuit yield optimization plays a vital role in ensuring reliable semiconductor manufacturing, directly impacting both product quality and production costs. Current approaches to yield optimization face two fundamental challenges that limit their practical effectiveness. First, yield estimation requires intensive computational resources. Second, traditional black-box optimization methods inefficiently allocate these resources across design candidates. Most existing approaches compound these issues by performing detailed yield estimations uniformly across all candidates, regardless of their potential quality. To address these limitations, we introduce a novel precision-aware yield optimization framework that intelligently adapts computational resource allocation based on each design candidate’s predicted performance. Our approach moves beyond simple simulation counting by incorporating a Figure of Merit (FoM) as a continuous quality metric. By combining a Continuous AutoRegression model to characterize the relationship between true yield and precision levels with a sophisticated multi-fidelity acquisition strategy, our framework achieves optimal resource distribution. Experimental validation on four industry-standard benchmark circuits demonstrates that our method converges with fewer than 1,000 simulations, reducing simulation costs by over $10 \times$ while achieving better final designs and robustness than state-of-the-art high-fidelity approaches.

Crossbar-based computing-in-memory (CIM) systems facilitate large-scale parallel multiply-and-accumulate (MAC) operations, while a domain-specific compiler (DSC) plays a pivotal role in optimizing the deployment of neural network algorithms on such systems. With the development of multi-core and large-core architectures, some key compiler problems such as high parallel processing, resource utilization, and crossbar array assignment methods have not been solved. For low-latency application scenarios, we have designed a resource scheduling strategy for our hardware system based on stream data processing to reduce the latency caused by intra-core and intercore communication. Additionally, a weight mapping strategy has been developed to maximize the potential of crossbar arrays in convolutional neural networks (CNNs) deployment. Experimental results on our multi-core eFlash-based CIM system-on-chip (SoC) demonstrate that these two technologies help CNNs achieve a 76% reduction in latency, a 30% improvement in resource utilization, and the use rate of crossbar array that can reach up to 94.7%.

Semiconductor yield estimation presents a critical challenge in modern manufacturing, directly impacting production costs and market competitiveness. Traditional estimation methods, particularly Monte Carlo simulation, while reliable, become computationally prohibitive for complex modern circuits. Contemporary approaches, including importance sampling and machine learning techniques, face fundamental limitations in consistency across circuit topologies and practical validation. This work introduces YieldAgent, a novel Large Language Model (LLM)-powered framework that revolutionizes yield estimation through dynamic integration of multiple analytical strategies. YieldAgent employs a three-layer agent architecture to analyze circuit characteristics and historical data, optimizing estimation methods while balancing computational efficiency and precision. The framework incorporates Retrieval-Augmented Generation for domain knowledge integration and Tree-structured Parzen Estimators for dynamic hyperparameter optimization. Experimental validation across 12nm and 40nm technology nodes demonstrates that YieldAgent reduces computational overhead by up to $2.9 \times$ while maintaining or exceeding state-of-the-art accuracy. The system’s ability to adapt across different circuit topologies and technology nodes establishes a new paradigm for scalable, intelligent yield estimation in electronic design automation.

As an emerging human-solving paradigm, crowdsourcing has attracted much attention where requesters want to employ reliable workers to complete the specific task. Task assignment is a vital branch in crowdsourcing. Most existing works in crowdsourcing haven't taken self-interested individuals' strategy into account. To guarantee truthfulness, auction has been regarded as a promising form to charge requesters and reward workers. In this paper, we consider an online task assignment scenario, where each worker has a set of experienced skills, whereas specific task is budget-constrained and requires certain skill. Under this scenario, we model the crowdsourcing task assignment as a reverse auction in which requesters are buyers and workers are sellers. Specially, our paper studies simple task case where the requester ask for single skill. We propose TMC-VCG and TMC-ST and prove the related properties for the mechanisms theoretically. Meanwhile, through extensive simulations, we verify the truthfulness and also evaluate other performance.

Auction is a highly effective trading form for distributing resource among buyers in a market at competitive prices, and has been applied to many domains, e.g. spectrum allocation in wireless networks and the virtual machine allocation in cloud computing. Most of existing auction mechanisms based on McAfee double auction calculate the uniform clearing prices for winning buyers no matter what channel they acquired, which does not reflect the differences of buyers' personalised preferences for heterogeneous spectrums. Hence, in this paper, we propose a truthful double auction scheme, named TRUDA, which incorporates the marginal effect of buyer–seller pair into auction mechanism design and considers the case in which buyers are mutually exclusive. We show analytically that this auction mechanism guarantees the economic‐robustness of the auction and has polynomial time complexity.

Space-time block code (STBC) has a good BER performance. In 3GPP FDD specifications, space time transmit diversity (STTD) employs STBC to maintain orthogonality between the two antennas in order to avoid self-interference in fading channels. However, the capacity of multiuser STBC systems is still severely affected by multiple access interference (MAI) and interpath interference (IPI). Fortunately, nonlinear parallel interference cancellation (NPIC) can effectively reduce MAI and IPI with simple signal processing. And for the small spreading factor systems, weighted NPIC (WNPIC) technique provides a good solution. In order to alleviate the influence of the MAI and IPI of multiuser STBC systems under the frequency-selective Rayleigh fading channel, WNPIC based space-time block code (WNPIC based STBC) scheme is proposed firstly in this paper. And then the error probability of the WNPIC based STBC is analyzed theoretically. From our analyses, RAKE based STBC can be regarded as 0th stage WNPIC based STBC whose weighted factors are one. Later, to further evaluate the proposed scheme and verify our analyses, some simulation resulted are achieved. These simulation results show us the optimum weighted factors under different user environments and the multiuser BER performance of simple (RAKE) STBC systems, NPIC based STBC, WNPIC based STBC systems.

As an important space-time code, the layered space-time (LST) code has been studying widely since it was firstly proposed by Foschini in 1996. To exploit its potential, the Bell Lab Layered Space-Time (BLAST) structure of the LST was proposed by Bell Lab. There are two types of BLAST architectures: vertical BLAST (V-BLAST) and diagonally BLAST (D-BLAST). Detection algorithms of V-BLAST were proposed by Golden (see Electronics Letters, vol.35, no.1, 1999). His detection process uses linear combination nulling and successive symbol cancellation (SSC) based on inversing and ordering. However, inversing (Moore-Penrose pseudoinverse) and ordering operation for each iteration bring a huge computation complexity. To detect M transmit antenna signals, Golden's detection algorithm needs M inversing and M ordering operations. Aiming at this shortcoming, a new detection algorithm for layered space-time code is proposed. This new sub-optimal detection algorithm is based on Greville inversing process, with two inversing and one ordering process. Simulation results show us the proposed sub-optimal scheme still has a good BER performance.

Space-time code techniques have been widely studied in wireless communications, for they provide better signal qualities in fading channel environments and increase the system capacity. As an important space-time code, Bell Lab Layered Space-Time (BLAST) code has been paid more attention. However, the traditional detection algorithm of V-BLAST space-time code needs much time to perform linear combination nulling and successive symbol cancellation. The greater the number of transmit antennas, the greater the time delay. To overcome this disadvantage, a new scheme based on linear combination nulling and parallel symbol cancellation is proposed. The new scheme has a lower time delay. The simulation results of this new scheme are obtained by COSSAP simulation in flat Rayleigh channels. These simulation results are presented to verify the BER performance of the new scheme based on parallel symbol cancellation.

In 3GPP FDD specifications, space time transmit diversity (STTD) employs space-time block code (STBC) to maintain orthogonality between the two antennas in order to avoid self-interference in fading channels. However, the STTD technique is just open loop transmit diversity, and we investigate whether we can obtain further gain if closed loop STBC is used. A novel scheme based on closed loop STBC is proposed first. And then an optimized allocation scheme of a fixed transmitter power budget between two transmit antennas is presented. Finally, COSSAP simulation results show a significant performance improvement between closed loop STBC and traditional STBC. Moreover, these simulation results also show a good agreement with our theoretical analyses.