Prompting techniques 13 General prompting / zero shot 13 One-shot & few-shot 15 System, contextual and role prompting 18 System prompting 19 Role prompting 21 Contextual prompting 23 Table of contents Design with simplicity 55 Be specific about the output 56 Use Instructions over Constraints 56 Control the max token length 58 Use variables in prompts 58 Experiment with input formats and writing styles need to figure out the model configuration. Most LLMs come with various configuration options that control the LLM’s output. Effective prompt engineering requires setting these configurations optimally for

multimodality across audio, visual, & text inputs 7/24: Apple releases Apple Intelligence, an AI system integrated into its devices, for developers 12/24: OpenAI announces o3, its highest-ever Unprecedented41 AI Performance = In 2024… Surpassed Human Levels of Accuracy & Realism, per Stanford HAI AI System Performance on MMLU Benchmark Test – 2019-2024, per Stanford HAI Note: The MMLU (Massive Multitask Human-Generated – 3/25, per Cameron Jones / Benjamin Bergen Date Released 5/24 1/25 2/25 AI system performance consistently improving over time AI Development Trending = Unprecedented43 AI Performance

a code change, 
 or generating a report. Applications that integrate LLMs but don’t use them to control workflow execution—think simple chatbots, single-turn LLMs, or sentiment classifiers—are not agents proactively correct its actions if needed. In case 
 of failure, it can halt execution and transfer control back to the user. 02 It has access to various tools to interact with external systems—both to gather agents. Well-documented, thoroughly tested, and reusable tools improve discoverability, simplify version management, and prevent redundant definitions. Broadly speaking, agents need three types of tools:


 a good fit. The Indeed team tested the previous job matching engine against the GPT-powered version with the new, customized context. 
 The performance uplift was significant: A 20% increase in job thousands of tasks every month, freeing people to do more high-impact work. Not surprisingly, the system is now spreading across other departments. It happened because we set bold automation goals from at a glance For our enterprise customers, nothing is more important than security, privacy and control. 
 Here’s how we ensure it: Your data stays yours We don’t use your content to train our models;

cost. As we employ expert parallelism during training, we also devise supplementary mechanisms to control communication overheads and ensure load balance. By combining these two techniques, DeepSeek-V2 features linear computations across different experts. In addition, MLA is also optimized based on an improved version of FlashAttention-2 (Dao, 2023). We conduct all experiments on a cluster equipped with NVIDIA H800 comprising 1.2M instances for helpfulness and 0.3M instances for safety. In comparison to the initial version, we improve the data quality to mitigate hallucinatory responses and enhance writing proficiency

dynamism ● Control flow (if, loop, etc) ● Dynamic shapes ○ Dynamic inputs: batch size, image size, sequence length, etc. ○ Output shape of some ops are data dependent: arange, nms, etc. ○ Control flow: as conv2d_NCHWc. Graph tuning is well defined for each subgraph. 3. Avoid runtime layout tracking system for operator requires layout transformation to optimize.© 2019, Amazon Web Services, Inc. or its

different integer division modes, floor division and truncating division. e Unified Object and Node system for TVM runtime o Lays groundwork forimproved multi-language support for expPosing runtime, and discuss more details at TVMConf. Oo oo QQ octoML 11 VM Memory Planning e Recently shipped a first version fn enain(0) -> Tensor[tk,)，f32] { ofdynamicmemory Planmng 寺中 竹2 o_ https:/github.com/apachei

更多 Deepseek 和 AI 资料，欢迎关注微信公众号【星禾光年 AI】，回复【deepseek】获取 1. 万能提示词生成模版：根据用户需求，帮助生成高质量提示词 SYSTEM 你是一位大模型提示词生成专家，请根据用户的需求编写一个智能助手的提示词，来指导大模型进行内容生成， 要求： 1. 以 Markdown 格式输出 2. 贴合用户需求，描述智能助手的定位、能力、知识储备 3 提示词应清晰、精确、易于理解，在保持质量的同时，尽可能简洁 4. 只输出提示词，不要输出多余解释 USER “ 请帮我生成一个 Linux ” 助手 的提示词 2. 文案大纲生成：根据用户提供的主题，来生成文案大纲 SYSTEM 你是一位文本大纲生成专家，擅长根据用户的需求创建一个有条理且易于扩展成完整文章的大纲，你拥有强大的 主题分析能力，能准确提取关键信息和核心要点。具备丰富的文案写作知识储备，熟悉各种文体和题材的文案大 创意性标题：为文章构思一个引人注目的标题，确保它既反映了文章的核心内容又能激发读者的好奇心。 USER “ ” 请帮我生成 中国农业情况 这篇文章的大纲 3. 中英翻译专家：中英文互译，对用户输入内容进行翻译 SYSTEM 你是一个中英文翻译专家，将用户输入的中文翻译成英文，或将用户输入的英文翻译成中文。对于非中文内容， 它将提供中文翻译结果。用户可以向助手发送需要翻译的内容，助手会回答相应的翻译结果，并确保符合中文语

How Would That Look Like?© 2019, Amazon Web Services, Inc. or its Affiliates. All rights reserved. System Overview Relay IR Graph Annotation with Your Annotator Graph Partitioning Your Codegen LLVM Inc. or its Affiliates. All rights reserved. Partition the Relay IR graph ● No user involvement System Overview Relay IR Graph Annotation with Your Annotator Graph Partitioning Your Codegen LLVM Inc. or its Affiliates. All rights reserved. Partition the Relay IR graph ● No user involvement System Overview Relay IR Graph Annotation with Your Annotator Graph Partitioning Your Codegen LLVM

speedup Engineering intensiveMachine Learning based Program Optimizer TVM: Learning-based Learning System High-level data flow graph and optimizations Directly generate optimized program for new operator Verilog VerilatorToward Unified IR InfraOverview of New IR Infra Single unified module/pass, type system, with function variants supportCompilation Flow under the New Infra IRModule (relay::Function)