大型语言模型(LLM)已成为推进Text-to-SQL任务的强大工具。研究发现,基于LLM的模型在不同评估指标下,其性能表现与经过微调的模型存在显著差异。因此,文章分析了测试套件执行准确度(EXE)和精确集匹配准确度(ESM)在评估基于LLM的Text-to-...大型语言模型(LLM)已成为推进Text-to-SQL任务的强大工具。研究发现,基于LLM的模型在不同评估指标下,其性能表现与经过微调的模型存在显著差异。因此,文章分析了测试套件执行准确度(EXE)和精确集匹配准确度(ESM)在评估基于LLM的Text-to-SQL模型时的不足,并提出了改进指标EESM(Enhanced Exact Set Matching)。实验结果表明,EXE和ESM分别存在高达13.2%和10.8%的假阳性和假阴性率,而EESM的假阳性率和假阴性率分别仅为0.2%和1.8%,表明EESM能够提供更准确的评估。展开更多
通过业务自适应组播路由和组播路由树的重构的方法来分析IPv6报文的QoS组播路由算法,分析了一种基于IPv6的QoS(quality of service)控制组播路由机制。仿真试验表明,该种路由算法是可行的,能够满足IPv6网络对相应QoS的要求,并具有很强...通过业务自适应组播路由和组播路由树的重构的方法来分析IPv6报文的QoS组播路由算法,分析了一种基于IPv6的QoS(quality of service)控制组播路由机制。仿真试验表明,该种路由算法是可行的,能够满足IPv6网络对相应QoS的要求,并具有很强的接收性能,在时间上也有较强的符合性。展开更多
In this paper, we analyze the physical layer abstraction for bit interleaved coded orthogonal frequency division multiplexing(BIC-OFDM) system from a parallel bit channel perspective. By combining the exponential effe...In this paper, we analyze the physical layer abstraction for bit interleaved coded orthogonal frequency division multiplexing(BIC-OFDM) system from a parallel bit channel perspective. By combining the exponential effective SNR(signal-to-noise ratio) mapping(EESM) with the maximum a posteriori(MAP) algorithm, a bit LLR(log-likelihood ratio) wise EESM(BL-EESM) method is proposed. This method can abstract the link performance with high accuracy, especially for the case when channel estimation is imperfect. Afterward, the BL-EESM method is simplified by utilizing the non-linear quantization idea, which can reduce the times of exponential operation by two orders of magnitude at wide system bandwidth, yet shows little loss in accuracy. Our proposal can be applied to both system level simulations to save the time consumption and to practical terminals to facilitate the adaptive modulation and coding(AMC) procedure, bringing about throughput improvement at low hardware cost.展开更多
In the future the fifth generation( 5 G) communication systems,channel models may be very complicated and it is difficult to calculate equivalent signal to interference plus noise ratio( SINR)of a random fading channe...In the future the fifth generation( 5 G) communication systems,channel models may be very complicated and it is difficult to calculate equivalent signal to interference plus noise ratio( SINR)of a random fading channel. Therefore,methods for the calculation of equivalent SINR of a random fading channel are very necessary.In this paper,an enhanced algorithm on the exponential effective SINR mapping( EESM) model for random fading channels was proposed. First, the optimal adjustment parameters of typical channel fading models including extended pedestrian A( EPA)model,extended vehicular A( EVA) model and extended typical urban( ETU) model were obtained by simulation. Then the proposed solution was used to actualize channel classification according to the maximum multipath delay and the average power of the random channel. The solution can determine the typical channel closest to random channel for obtaining the optimal adjustment value of EESM. The evaluation results indicate that the proposed one can improve the whole system throughput significantly and meanwhile the accuracy of the link prediction algorithm is also guaranteed.展开更多
文摘大型语言模型(LLM)已成为推进Text-to-SQL任务的强大工具。研究发现,基于LLM的模型在不同评估指标下,其性能表现与经过微调的模型存在显著差异。因此,文章分析了测试套件执行准确度(EXE)和精确集匹配准确度(ESM)在评估基于LLM的Text-to-SQL模型时的不足,并提出了改进指标EESM(Enhanced Exact Set Matching)。实验结果表明,EXE和ESM分别存在高达13.2%和10.8%的假阳性和假阴性率,而EESM的假阳性率和假阴性率分别仅为0.2%和1.8%,表明EESM能够提供更准确的评估。
文摘通过业务自适应组播路由和组播路由树的重构的方法来分析IPv6报文的QoS组播路由算法,分析了一种基于IPv6的QoS(quality of service)控制组播路由机制。仿真试验表明,该种路由算法是可行的,能够满足IPv6网络对相应QoS的要求,并具有很强的接收性能,在时间上也有较强的符合性。
基金the Shanghai Basic Research KeyProject(No.11DZ1500206)the NationalScience and Technology Major Project of China(No.2012ZX03001013-003)
文摘In this paper, we analyze the physical layer abstraction for bit interleaved coded orthogonal frequency division multiplexing(BIC-OFDM) system from a parallel bit channel perspective. By combining the exponential effective SNR(signal-to-noise ratio) mapping(EESM) with the maximum a posteriori(MAP) algorithm, a bit LLR(log-likelihood ratio) wise EESM(BL-EESM) method is proposed. This method can abstract the link performance with high accuracy, especially for the case when channel estimation is imperfect. Afterward, the BL-EESM method is simplified by utilizing the non-linear quantization idea, which can reduce the times of exponential operation by two orders of magnitude at wide system bandwidth, yet shows little loss in accuracy. Our proposal can be applied to both system level simulations to save the time consumption and to practical terminals to facilitate the adaptive modulation and coding(AMC) procedure, bringing about throughput improvement at low hardware cost.
基金Institute of Nonlinear Science of Donghua University,China
文摘In the future the fifth generation( 5 G) communication systems,channel models may be very complicated and it is difficult to calculate equivalent signal to interference plus noise ratio( SINR)of a random fading channel. Therefore,methods for the calculation of equivalent SINR of a random fading channel are very necessary.In this paper,an enhanced algorithm on the exponential effective SINR mapping( EESM) model for random fading channels was proposed. First, the optimal adjustment parameters of typical channel fading models including extended pedestrian A( EPA)model,extended vehicular A( EVA) model and extended typical urban( ETU) model were obtained by simulation. Then the proposed solution was used to actualize channel classification according to the maximum multipath delay and the average power of the random channel. The solution can determine the typical channel closest to random channel for obtaining the optimal adjustment value of EESM. The evaluation results indicate that the proposed one can improve the whole system throughput significantly and meanwhile the accuracy of the link prediction algorithm is also guaranteed.
