金融界2025年8月12日消息，国家知识产权局信息显示，腾讯科技（深圳）有限公司取得一项名为"车轮 触地点的获取方法、存储介质和电子设备"的专利，授权公告号CN114627069B，申请日期为2022年03 月。 本文源自：金融界 作者：情报员 天眼查资料显示，腾讯科技（深圳）有限公司，成立于2000年，位于深圳市，是一家以从事软件和信息 技术服务业为主的企业。企业注册资本200万美元。通过天眼查大数据分析，腾讯科技（深圳）有限公 司共对外投资了15家企业，参与招投标项目264次，财产线索方面有商标信息5000条，专利信息5000 条，此外企业还拥有行政许可533个。 ...

Group 1 - Shenzhen Xunlu Innovation Technology Co., Ltd. recently completed its Series A financing round, raising tens of millions of RMB, with investors including Sequoia China Seed Fund and Shunwei Capital [1] - The company was established in 2023 and is primarily engaged in software and information technology services, with a registered capital of 1.808079 million RMB [1] - As of 2025, the company has completed its Series A financing round, with the transaction amount being tens of millions of RMB [1] Group 2 - Shenzhen Xunlu Innovation Technology Co., Ltd. holds 13 trademark registrations and 34 patents, along with 3 administrative licenses [1] - The shareholders of the company include Tarand Tech Limited, Astrend V (Hong Kong) Beta Limited, Dexun (Shenzhen) Global Venture Capital Co., Ltd., Shenzhen Milu Enterprise Management Partnership (Limited Partnership), and Shenzhen Lulul Enterprise Management Partnership (Limited Partnership) [1]

Group 1 - Suzhou Puyuan Smart Technology Co., Ltd. has been established with Dai Fengqiang as the legal representative [1] - The company is wholly owned by Puyuan Information Technology Co., Ltd. (688118) [1] - The registered capital of Suzhou Puyuan Smart Technology Co., Ltd. is 5 million yuan [2] Group 2 - The business scope includes wholesale and retail of computer software and hardware, software sales, and information system integration services [1][2] - The company is located in Suzhou High-tech Zone, Jiangsu Province [2] - The business registration is valid until August 7, 2025, with no fixed end date [2]

专利摘要显示，本申请公开了一种大模型推理方法、装置、相关设备及计算机程序产品，涉及人工智能 领域，本申请先获取待处理问题的问题陈述，而后在大模型基于问题陈述进行链式推理过程中，依据当 前推理节点的输入token判断当前节点是否为关键节点，若不是，则对前节点的原始输出token进行压缩 处理，基于压缩处理结果推理后续节点。本申请先对推理节点关键与否进行了判定，而后在进行后续推 理时压缩非关键推理节点的输出token，减少了后续推理过程所需处理的输入数据量，加快了后续处理 速度，有助于优化链式推理，特别是长链式推理的计算效率。 金融界2025年8月6日消息，国家知识产权局信息显示，科大讯飞股份有限公司申请一项名为"大模型推 理方法、装置、相关设备及计算机程序产品"的专利，公开号CN120430418A，申请日期为2025年07 月。 天眼查资料显示，科大讯飞股份有限公司，成立于1999年，位于合肥市，是一家以从事软件和信息技术 服务业为主的企业。企业注册资本231173.4185万人民币。通过天眼查大数据分析，科大讯飞股份有限 公司共对外投资了129家企业，参与招投标项目5000次，财产线索方面有商标信息2744 ...

Group 1 - The core point of the article is that Tencent Technology (Shenzhen) Co., Ltd. has obtained a patent for a method and device related to "inference engine invocation," with the patent granted on August 5, 2025, and the application date being March 2022 [1] - Tencent Technology was established in 2000 and is primarily engaged in software and information technology services, with a registered capital of 2 million USD [1] - The company has invested in 15 enterprises and participated in 262 bidding projects, holding 5,000 trademark records and 5,000 patent records, along with 527 administrative licenses [1]

博汇科技（SH 688004，收盘价：25.05元）8月4日晚间发布公告称，截至2025年7月31日，公司通过上 海证券交易所交易系统以集中竞价交易方式已累计回购公司股份约132万股，占公司总股本的比例为 1.6516%，回购成交的最高价为21.6元/股，最低价为13.43元/股，成交金额总额为人民币约2336万元。 （文章来源：每日经济新闻） 2024年1至12月份，博汇科技的营业收入构成为：软件和信息技术服务业占比100.0%。 ...

Group 1 - The core viewpoint of the news is that Cambricon Technologies Co., Ltd. (寒武纪) has shown significant financial growth in its latest quarterly report, with a total revenue of 1.111 billion yuan, representing a year-on-year increase of 4230.22% [1] - The company's net profit attributable to shareholders reached 355 million yuan, up 256.82% year-on-year, while the non-recurring net profit was 276 million yuan, reflecting a growth of 205.68% [1] - The company's liquidity ratios are strong, with a current ratio of 8.373 and a quick ratio of 4.495, indicating a solid financial position [1] Group 2 - Cambricon has made investments in 18 companies and participated in 35 bidding projects, showcasing its active engagement in the market [2] - The company holds a substantial intellectual property portfolio, including 820 trademarks and 752 patents, along with 5 administrative licenses [2] - Established in 2016 and based in Beijing, Cambricon focuses on software and information technology services, with a registered capital of approximately 4.183 billion yuan [1][2]

Core Viewpoint - The article introduces MixGRPO, a new framework that combines Stochastic Differential Equations (SDE) and Ordinary Differential Equations (ODE) to enhance the efficiency and performance of image generation processes [1][81]. Group 1: MixGRPO Framework - MixGRPO simplifies the optimization process in Markov Decision Processes (MDP) by utilizing a mixed sampling strategy, which improves both efficiency and performance [1][17]. - The framework shows significant improvements in human preference alignment across multiple dimensions, outperforming DanceGRPO with a training time reduction of nearly 50% [2][60]. - MixGRPO-Flash, a faster variant of MixGRPO, further reduces training time by 71% while maintaining similar performance levels [2][60]. Group 2: Performance Metrics - In comparative studies, MixGRPO achieved a higher Unified Reward score of 3.418, compared to DanceGRPO's 3.397, indicating better alignment with human preferences [60]. - MixGRPO-Flash demonstrated an average iteration time of 112.372 seconds, significantly lower than DanceGRPO's 291.284 seconds [60]. Group 3: Sampling Strategy - The MixGRPO framework employs a hybrid sampling method, where SDE sampling is used within a defined interval during the denoising process, while ODE sampling is applied outside this interval [14][20]. - This approach allows for a reduction in computational overhead and optimization difficulty, while ensuring that the sampling process remains aligned with the marginal distributions of SDE and ODE [30][81]. Group 4: Sliding Window Strategy - A sliding window strategy is introduced to optimize the denoising steps, allowing the model to focus on specific time steps during training [32][35]. - The research team identified key hyperparameters for the sliding window, including window size and movement intervals, which significantly impact performance [34][70]. Group 5: High-Order ODE Solvers - The integration of high-order ODE solvers, such as DPM-Solver++, enhances the sampling speed during the GRPO training process, balancing computational cost and performance [45][76]. - The experiments indicated that a second-order midpoint method was optimal for the high-order solver settings [76]. Group 6: Experimental Validation - The experiments utilized the HPDv2 dataset, which includes diverse prompts, demonstrating that MixGRPO can achieve effective human preference alignment with a limited number of training prompts [49][50]. - The results from various reward models confirmed the robustness of MixGRPO, showing superior performance in both single and multi-reward settings [56][82].

Group 1 - The core point of the article is that Wangsu Technology Co., Ltd. has obtained a patent for a "model training method, system, electronic device, and storage medium" with the authorization announcement number CN114239848B, applied for on November 2021 [1] - Wangsu Technology Co., Ltd. was established in 2000 and is located in Shanghai, primarily engaged in software and information technology services [1] - The registered capital of Wangsu Technology Co., Ltd. is 2,445.73 million RMB [1] Group 2 - According to data analysis from Tianyancha, Wangsu Technology Co., Ltd. has invested in 39 companies and participated in 1,125 bidding projects [1] - The company has 269 trademark information entries and 1,292 patent information entries, along with 6 administrative licenses [1]

Core Insights - The software and information technology services industry in China is experiencing robust growth, with both revenue and profit showing double-digit increases in the first half of 2025 [1] Revenue Performance - Software business revenue reached 70,585 billion yuan, marking an 11.9% year-on-year growth, indicating steady revenue growth alongside improved profitability [1] - Total profit from software business increased by 12% year-on-year, reflecting a positive development trend [1] Export Trends - Software exports amounted to 28.3 billion USD, with a year-on-year growth of 5.3%, demonstrating the industry's steady progress in expanding overseas markets [1] Sector Breakdown - Revenue from software products was 15,441 billion yuan, showing stable growth and accounting for 21.9% of the total industry revenue, providing strong support for the industry's stable operation [1] - Revenue from information technology services continued to grow at a double-digit rate, reaching 48,362 billion yuan, which constitutes nearly 70% of the total industry revenue, making it the main driver of industry growth [1]