定制报告-个性化定制-按需专项定制研究报告

行业报告、薪酬报告

联系：400-6363-638

《亚信科技&清华AIR：2025年DeepSeek赋能自智网络高阶演进评测报告（42页）.pdf》由会员分享，可在线阅读，更多相关《亚信科技&清华AIR：2025年DeepSeek赋能自智网络高阶演进评测报告（42页）.pdf（42页珍藏版）》请在薪酬报告网上搜索。

1、?1?DeepSeek?DeepSeek?2?(?)?DeepSeek?2025?2?Test Report-DeepSeek Empowers the Advanced Evolution of Autonomous Networks,Ye Ouyang,Yunxin Liu,et al,2025.2.?3?1.DeepSeek?.5 2.?L4?.5 3.?.6 4.?.6 5.DeepSeek?.6 5.1 测试目标.6 5.2 测试环境搭建.7 5.2.1?.7 5.2.2?.7 5.2.3?.7 5.3 DeepSeek 相关通用能力.7 5.3.1?.7 5.3.2?.8 5.3.

2、3?.9 5.3.4?.10 5.3.5?.11 5.3.6?.11 5.4 高价值测试场景选取.12 6.?.13 6.1 场景 1、业务编排智能设计.13 6.1.1?.13 6.1.2?.14 6.1.3?.14 6.2 场景 2、网络数据查询分析.16 6.2.1?.16 6.2.2?.17 6.2.3?.17 6.3 场景 3、网络拓扑生成.19 6.3.1?.19 6.3.2?.19 6.3.3?.20 6.4 场景 4、网络故障根因分析.21 6.4.1?.21 6.4.2?.22?4 6.4.3?.22 6.5 场景 5、IP 网配置生成.25 6.5.1 IP?.25 6.5

3、.2 IP?.25 6.5.3 IP?.26 6.6 场景 6、一线装维服务.27 6.6.1?.27 6.6.2?.28 6.6.3?.29 6.7 场景 7、感知诊断分析.30 6.7.1?.30 6.7.2?.31 6.7.3?.32 6.8 场景 8、无线投诉处理.33 6.8.1?.33 6.8.2?.34 6.8.3?.34 7.?.36 7.1 自智网络场景应用 DeepSeek 优势.37 7.2 DeepSeek 赋能自智网络场景不足.38 7.3 DeepSeek 赋能自智网络优化方向.38 8.DeepSeek?.39 9.?.39 10.?.41?5?DeepSeek?

4、V3?R1?AISWare?AN?Evo1?DeepSeek?V3?R1 模?DeepSeek?务?AN?Evo?DeepSeek?够?撰?DeepSeek?撰?告?DeepSeek?撑练撰?模?1.DeepSeek?DeepSeek?(AGI)?务?尽?DeepSeek?LLM?DeepSeek?LLM?社?空?LLM?DeepSeek-R1?DeepSeek-V3?,?撰?务?DeepSeek?撰?OpenAI?GPT?告?GPT-4?10%?DeepSeek?撰?撰?AI?DeepSeek?AI?2.?L4?L3?L4?DeepSeek?1AISWare AN Evo?AISWare A

5、utonomous Networks Evolution?6 3.?DeepSeek?务?确?研?撑练?4.?4-1?AISWare?AN?Evo?模?AN Evo?DeepSeek?务?5.DeepSeek?5.1?DeepSeek?务?L4/L5?研?l?务?DeepSeek?够?DeepSeek?务?l 模?务?7 l?副?导?DeepSeek?务?研?5.2?5.2.1?NVIDIA A800 80GB GPU?16?GPU?l?GA100?Ampere?l CUDA?6912?l?80GB HBM2e?l?1935GB/s?l?250W?l?PCIe 4.0 x16?5.2.2?l

6、CUDA?GPU?l vLLM?空?撰?l PyTorch?l Flash Attention?Transformer?l Transformers?撰?撰?5.2.3?5?l DeepSeek-R1 l DeepSeek-V3 l?务?3?C/D/E?C?5.3 DeepSeek?6?务?够?5.3.1?8 l 撰?1)?撰?务?副?2)?撰?研?副?3)?副?=?副?100%?=?100%?模?撰?务模?模?=?模?100%?撰?研?撑练?撑?l?1)?导?2)?3)?=?100%?=?100%?5.3.2?l?1)?务?研?2)?3)?9?=?100%?研?务?l?1)?务?2)?3)?

7、=?务?100%?=?务?100%?副?l?1)?2)?3)?=?100%?撑练?务?5.3.3?l?1)?研?2)?3)?研?10?=?100%?=?100%?l?1)?2)?3)?=?100%?撑练?务?5.3.4?够?研?务?1)?撑练?务?2)?务?研?够?3)?=?100%?=?100%?11?=?100%?研?撑练?务?够?5.3.5?研务?1)?研?务?2)?撰?撑练?撰?3)?模?=?100%?=?100%?=?100%?撰?撑练?5.3.6?研?务?1)?研?2)?撰?撰?3)?12?=?100%?撰?研?撰?撰?=?撰?100%?撰?5.4?5-1?5?够?TMF?撰?CS

8、Ps?副?13?5-2?AI?1?TMF?AI?务?Copilot?Agent?DeepSeek?AISWare AN Evo?务?8?DeepSeek?5-1?IP?6.?6.1?1?6.1.1?14?务?模?AI?6-1?1?撰?2?DeepSeek R1?DeepSeek V3?800?模?3?研?6.1.2?6-1?A?100Mbps?%?A?100Mbps?%?800?%?%?800?%?%?800?6.1.3?15 1)?研?5.3?5.3?5.3?93.40%90.20%93.00%90.10%90.20%92.90%91.60%91.60%92.60%92.60%94.90%9

9、4.40%92.00%93.40%95.40%80.81%85.20%89.20%89.20%89.20%80.81%84.70%84.70%85.70%84.70%业务智能编排语义连贯性业务智能编排准确性业务智能编排完整性业务智能编排识别率业务智能编排语义一致性?DeepSeek R1DeepSeek V3模型C模型D模型E95.00%96.00%91.20%89.00%89.00%91.60%93.00%94.00%94.40%90.20%90.20%80.81%84.70%84.70%70.00%业务智能编排推断合理性业务智能编排隐含信息理解业务智能编排因果关系准确性?DeepSeek

10、R1DeepSeek V3模型C模型D模型E90.20%90.00%89.00%91.60%91.50%88.00%94.40%93.00%85.00%80.81%90.20%79.00%70.00%84.70%76.00%业务智能编排准确性业务智能编排全面性业务智能编排时效性?DeepSeek R1DeepSeek V3模型C模型D模型E?16 2)?研?DeepSeek R1?DeepSeek V3?C?模?90%?副?D?E?DeepSeek R1?DeepSeek V3?C?模?90%?副?够?撑练?务?撰?模?D?E?撰?撑练?DeepSeek R1?DeepSeek V3?C?90

11、%?副?务?D?E?务?副?DeepSeek R1?告?6-2?DeepSeek R1 DeepSeek V3?C?D?E?3)?DeepSeek?C?务?DeepSeek?R1?告?a)?u DeepSeek R1?V3?b)?u DeepSeek-R1?6.2?2?6.2.1?17?研?撰?务?研?研?模?模?研?6-2?1.?2.?研?3.?DeepSeek R1/DeepSeek V3?研?6.2.2?6-3?%?%?1200?%?%?1200?6.2.3?1)?研?研?研?5.3?18?研?5.3?2)?研?研?DeepSeek R1?DeepSeek V3?研?90%?副?模?C?

12、告?D?E?撑练?DeepSeek R1?DeepSeek V3?6-4?DeepSeek R1 DeepSeek V3?C?D?E?95.40%93.20%95.00%90.10%90.20%94.90%91.60%91.60%92.60%92.60%94.90%94.40%92.00%93.40%95.40%80.81%89.20%90.20%89.20%89.20%80.31%84.70%84.70%85.70%84.70%网络数据查询分析语义连贯性网络数据查询分析准确性网络数据查询分析完整性网络数据查询分析识别率网络数据查询分析语义一致性?研?研?DeepSeek R1DeepSeek

13、 V3模型C模型D模型E95.87%99.60%94.00%94.09%94.90%91.60%94.78%97.90%94.40%80.81%80.81%80.81%70.00%76.90%70.00%网络数据查询分析意图识别准确率网络数据查询分析提取准确率网络数据查询分析覆盖范围?研?研?DeepSeek R1DeepSeek V3模型C模型D模型E?19 3)?研?DeepSeek R1?DeepSeek V3?研?务?研?C?告?D?E?撑练?a)?u DeepSeek R1?V3?模?u DeepSeek R1?V3?b)?u DeepSeek-R1?撰?够?25?6.3?3?6.3

14、.1?务?AI?6-3?1.?2.?研?3.?DeepSeek R1/DeepSeek V3?研?6.3.2?6-5?10?20?%?务?1?2?JSON?%?600?A?%?务?A?JSON?%?600?6.3.3?1)?研?5.3?5.3?94.00%90.20%92.00%92.00%90.50%94.50%90.60%92.10%91.10%92.60%97.50%94.90%92.50%93.50%95.40%80.31%91.20%90.20%85.20%89.20%85.90%85.70%84.70%85.70%84.70%网络拓扑生成语义连贯性网络拓扑生成准确性网络拓扑生成完整

15、性网络拓扑生成识别率网络拓扑生成语义一致性?DeepSeek R1DeepSeek V3模型C模型D模型E?21 2)?研?DeepSeek?R1?DeepSeek?V3?C?告?D?E?DeepSeek?R1?DeepSeek?V3?6-6?DeepSeek R1 DeepSeek V3?C?D?E?3)?DeepSeek?模?C?DeepSeek?C?a)?u DeepSeek R1?V3?b)?u?DeepSeek R1?撰?够?2?5?6.4?4?6.4.1?务?模?够?95.87%94.40%91.20%94.09%94.90%91.60%94.78%97.90%94.40%80.8

16、1%80.81%80.81%70.00%96.90%70.00%网络拓扑生成意图识别准确率网络拓扑生成提取准确率网络拓扑生成覆盖范围?DeepSeek R1DeepSeek V3模型C模型D模型E?22?6-4?1.?2.?研?3.?DeepSeek R1/DeepSeek V3?副?研?6.4.2?6-7?%?%?1500?%?%?1500?%?%?1500?%?%?1500?6.4.3?23 1)?研?5.3?够?5.3?5.3?5.3?95.87%94.40%91.20%94.09%94.90%91.60%94.78%97.90%94.40%80.81%80.81%80.81%70.00

17、%96.90%70.00%网络故障根因分析意图识别准确率网络故障根因分析提取准确率网络故障根因分析覆盖范围?DeepSeek R1DeepSeek V3模型C模型D模型E90.20%90.00%89.00%91.60%91.50%88.00%94.40%93.00%85.00%80.81%90.20%79.00%70.00%84.70%76.00%网络故障根因分析诊断准确性网络故障根因分析方案合理性网络故障根因分析时效性?DeepSeek R1DeepSeek V3模型C模型D模型E90.20%90.00%89.00%91.60%91.50%88.00%94.40%93.00%85.00%94

18、.40%90.20%79.00%80.00%84.70%76.00%网络故障根因分析准确性网络故障根因分析全面性网络故障根因分析时效性?DeepSeek R1DeepSeek V3模型C模型D模型E?24 2)?研?DeepSeek?R1?DeepSeek?V3?C?D?E?够?DeepSeek?R1?DeepSeek?V3?C?D?够?E?DeepSeek?R1?DeepSeek?V3?务?90%?C?D?E?DeepSeek?R1?DeepSeek?V3?研?C?D?E?6-8?DeepSeek R1 DeepSeek V3?C?D?E?3)?DeepSeek?R1?DeepSeek?V3

19、?够?90%?副?a)?u DeepSeek R1?V3?DeepSeek R1?务?95.00%96.00%91.20%89.00%89.00%91.60%93.00%94.00%94.40%90.20%90.20%80.81%84.70%84.70%70.00%网络故障根因分析推断合理性 网络故障根因分析隐含信息理解网络故障根因分析因果关系准确性?DeepSeek R1DeepSeek V3模型C模型D模型E?25?够?C?b)?u DeepSeek-R1?6.5?5?IP?6.5.1 IP?IP?务?IP?模?IP?IP?6-5 IP?1.?IP?IP?2.?研?3.?DeepSeek

20、R1/DeepSeek V3?副?研?6.5.2 IP?6-9 IP?IP?100?IP?%?100?IP?255.255.255.0?eth0?%?IP?200?C?%?C?%?IP?200?26?D?%?D?模?%?IP?200?6.5.3 IP?1)IP?研 IP?5.3?IP?5.3?IP?5.3?95.87%94.40%91.20%94.09%94.90%91.60%94.78%97.90%94.40%80.81%80.81%80.81%70.00%96.90%70.00%IP网配置生成意图识别准确率IP网配置生成提取准确率IP网配置生成覆盖范围IPIP?DeepSeek R1Dee

21、pSeek V3模型C模型D模型E90.20%90.00%89.00%91.60%91.50%88.00%94.40%93.00%85.00%80.81%90.20%79.00%70.00%84.70%76.00%IP网配置生成准确性IP网配置生成全面性IP网配置生成时效性IPIP?DeepSeek R1DeepSeek V3模型C模型D模型E?27 2)?研?IP?DeepSeek?R1?DeepSeek?V3?C?告?D?E?IP?DeepSeek?R1?DeepSeek?V3?IP?务?6-10?DeepSeek R1 DeepSeek V3?C?D?E?3)IP?DeepSeek?R1

22、?DeepSeek?V3?91%?副?95%?C?DeepSeek?R1?C?副?研?IP?a)?u DeepSeek R1 撑练?b)?u DeepSeek-R1?6.6?6?6.6.1?95.00%96.00%89.00%89.00%93.00%94.00%90.20%90.20%84.70%84.70%IP网配置生成相关性IP网配置生成多样性IPIP?DeepSeek R1DeepSeek V3模型C模型D模型E?28?务?模?务?6-6?1.?2.?研?3.?DeepSeek R1/DeepSeek V3?副?研?6.6.2?6-11?ONU CPU?ONU?务?务?映?JSON?12

23、00?ONU CPU?ONU?1200?务?1200?29?FTTH?DSL?Cable Modem?4G/5G?6.6.3?1)?研?5.3?5.3?5.3?95.87%94.40%91.20%94.09%94.90%91.60%94.78%97.90%94.40%80.81%80.81%80.81%70.00%96.90%70.00%一线装维服务意图识别准确率一线装维服务提取准确率一线装维服务覆盖范围?DeepSeek R1DeepSeek V3模型C模型D模型E93.40%90.20%90.00%90.00%90.00%94.10%90.60%91.60%92.50%91.60%97.5

24、0%93.40%92.00%95.00%94.00%80.61%80.20%88.20%88.20%88.20%80.00%80.70%85.70%85.70%85.70%一线装维服务语义连贯性一线装维服务准确性一线装维服务完整性一线装维服务识别率一线装维服务语义一致性?DeepSeek R1DeepSeek V3模型C模型D模型E?30 2)?研?DeepSeek?R1?DeepSeek?V3?94%?副?C?告?D?E?DeepSeek?R1?DeepSeek?V3?DeepSeek?R1?DeepSeek?V3?C?D?务?E?6-12?DeepSeek R1 DeepSeek V3?C

25、?D?E?3)?务?DeepSeek?R1?DeepSeek?V3?C?D?务?E?a)?u DeepSeek R1 撑练?b)?u?6.7?7?6.7.1?92.20%93.20%80.00%92.60%95.60%85.00%92.40%95.40%94.00%91.00%92.00%91.00%88.00%89.00%88.00%一线装维服务准确性一线装维服务全面性一线装维服务时效性?DeepSeek R1DeepSeek V3模型C模型D模型E?31?6-7?1.?2.?研?3.?DeepSeek R1/DeepSeek V3?副?研?6.7.2?6-13?13075649799?20

26、23?3?12?2023?3?18?务?务?JSON?1000?务?1000?务?1000?32?6.7.3?1)?研?5.3?5.3?5.3?95.80%93.40%90.20%94.00%93.90%90.40%94.88%97.60%94.30%80.10%80.81%80.81%74.00%76.00%70.00%感知诊断分析意图识别准确率感知诊断分析提取准确率感知诊断分析覆盖范围感知诊断分析指标DeepSeek R1DeepSeek V3模型C模型D模型E94.40%91.20%89.00%91.00%89.00%94.50%91.60%93.60%92.50%92.60%97.50

27、%93.64%93.00%95.00%93.00%81.61%81.20%88.20%88.20%88.20%80.00%81.70%85.70%85.70%85.70%感知诊断分析语义连贯性感知诊断分析准确性感知诊断分析完整性感知诊断分析识别率感知诊断分析语义一致性?DeepSeek R1DeepSeek V3模型C模型D模型E?33 2)?研?DeepSeek?R1?DeepSeek?V3?C?告?D?E?DeepSeek?R1?DeepSeek?V3?务?DeepSeek?R1?6-14?DeepSeek R1 DeepSeek V3?C?D?E?3)?务?DeepSeek?R1?Dee

28、pSeek?V3?C?D?务?a)?u DeepSeek R1?撑练?撰?撑练?b)?u DeepSeek R1?撑练?6.8?8?6.8.1?92.40%92.20%80.00%92.70%94.60%85.00%94.40%95.40%94.00%91.00%91.00%90.00%88.00%89.00%88.00%感知诊断分析准确性感知诊断分析全面性感知诊断分析时效性?DeepSeek R1DeepSeek V3模型C模型D模型E?34?务?导?研?6-8?1?2?研?3?DeepSeek R1/DeepSeek V3?副?研?6.8.2?6-15?394198-71?2546248-

29、61?%?%?1000?%?%?1000?6.8.3?1)?研?35?5.3?5.3?2)?研?DeepSeek?R1?DeepSeek?V3?C?告?D?E?DeepSeek?R1?DeepSeek?V3?6-16?DeepSeek R1 DeepSeek V3?C?D?E?3)?95.87%90.87%89.00%94.09%93.10%88.00%94.78%91.00%85.00%80.81%79.00%79.00%70.00%70.00%76.00%无线网络优化语义连贯性无线网络优化准确性无线网络优化完整性?DeepSeek R1DeepSeek V3模型C模型D模型E94.40%9

30、0.20%90.00%90.10%90.20%94.90%91.60%91.60%92.60%92.60%97.90%94.40%92.00%93.40%95.40%80.81%90.20%90.20%89.20%89.20%96.90%84.70%84.70%85.70%84.70%无线网络优化语义连贯性无线网络优化准确性无线网络优化完整性无线网络优化识别率无线网络优化语义一致性?DeepSeek R1DeepSeek V3模型C模型D模型E?36 DeepSeek?R1?DeepSeek?V3?C?DeepSeek?R1?DeepSeek?V3?C?D?a)?u DeepSeek R1?V

31、3?务?b)?u?撑练?覆?撰?撰?7.?够?撰?研?DeepSeek?DeepSeek?务?DeepSeek?务?7-1?DeepSeek R1 DeepSeek V3 C D E?IP?37?务?7.1?DeepSeek?DeepSeek?撰?务?够?研?DeepSeek?l?务?DeepSeek R1?V3?够?够?务?研?DeepSeek R1?V3?90%?副?C?务?告?撑练?够?DeepSeek R1?V3?l?IP?”?务?DeepSeek R1?V3?IP?100?IP?模?DeepSeek R1?V3?C?务?告?撑练?DeepSeek R1?V3?D?务?告?模?Deep

32、Seek R1?V3?l?务?DeepSeek R1?V3?研?DeepSeek R1?V3?94%?副?C?务?告?撑练?DeepSeek R1?V3?l?务?DeepSeek R1?V3?务?DeepSeek R1?V3?90%?副?38?副?DeepSeek R1?V3?务?务?DeepSeek R1?V3?C?务?7.2 DeepSeek?务?DeepSeek?撑练?DeepSeek R1?V3?“?”?“?”?务?撰?够?2?5?务?C?务?务?C?DeepSeek R1?V3?“?”?“?”?务?务?DeepSeek R1?V3?够?“?”?务?撑练?撰?7.3 DeepSeek?

33、DeepSeek?务?l?务?务?l?l?务?撑练?确?够?l?务?39 l?RAG?务?RAG?Retrieval-Augmented Generation?撑练?撰?“?”?“?”?务?RAG?8.DeepSeek?务?务?DeepSeek?l?DeepSeek?够?模?C?l?DeepSeek?l?DeepSeek?C?l?DeepSeek?l?务?RAG?Retrieval-Augmented Generation?DeepSeek?务?DeepSeek?务?够?撑练?9.?1 DeepSeek,DeepSeek v3:Technical advancements in multimo

34、dal reasoning and code synthesis,DeepSeek,Tech.Rep.DS-2024-003,2024.Online.Available:https:/ 10,2024.2 DeepSeek,DeepSeek-R1:Technical report on long-context MoE models,DeepSeek,Tech.Rep.DS-2023-001,2023.Online.Available:?40 https:/ 10,2024.3 DeepSeek.API documentation for DeepSeek-Lite model.(2024).

35、Online.Available:https:/ 10,2024.4“IG1401_TMForum_AN_Guide_level_4_industry_blueprint_v6.0.0”Nov.2024 TM Forum 5“TM_Forum_Autonomous_networks_Level_4_industry_blueprint”June.2024 TM Forum 6“IG1326_AN_Empowering_DT_Evolving_towards_Level_4_v1.0.0”June.2024 TM Forum 7“Autonomous Networks Industry Blue

36、print:An Evolution Guide towards L4”2024 TM Forum 8“TS 28.100 Management and orchestration;Levels of autonomous network”3GPP 9“TS 28.312 Management and orchestration;Intent driven Management services for mobile networks”,3GPP 10“Autonomous networks:empowering digital transformation for the telecoms industry”,2019 TM Forum 11“Hype Cycle for Operations and Automation in the Communications Industry”2024 Gartner 12?2024?务?13?务?2024?3?14?2024?2024 副?15?2025-02 务?.16?2024-06-18 ITU-T 17?2023-05-10 TM Forum.?41 10.?务?尽?务?尽?务?络?10?100193?010-82166699?010-82166688 Email?5G?Thank you?