{"id":11832,"date":"2023-11-02T19:00:15","date_gmt":"2023-11-02T11:00:15","guid":{"rendered":"https:\/\/ihower.tw\/blog\/?p=11832"},"modified":"2025-07-19T19:37:07","modified_gmt":"2025-07-19T11:37:07","slug":"aie-writing-principles-for-task-tuned-prompt-engineering","status":"publish","type":"post","link":"https:\/\/ihower.tw\/blog\/11832-aie-writing-principles-for-task-tuned-prompt-engineering","title":{"rendered":"AI Engineer \u9031\u5831\u00a0\ud83d\ude80 Writing Principles for Task-Tuned Prompt Engineering \u7b46\u8a18 #03"},"content":{"rendered":"\n
\"\"<\/a><\/figure>\n\n\n\n

Hello! \u4f60\u597d \ud83d\udc4b<\/p>\n\n\n\n

AI Engineer Summit \u5df2\u7d93\u904e\u53bb\u5feb\u4e00\u500b\u6708\uff0c\u5b98\u65b9\u9664\u4e86\u9678\u7e8c\u91cb\u51fa\u5f8c\u88fd\u9304\u5f71<\/a>\uff0c\u539f\u5148\u76f4\u64a5\u6c92\u6709\u7684 Workshop \u4e5f\u6709\u8cc7\u6599\u4e86\uff0c\u8b93\u6211\u5011\u4f86\u770b\u770b\u3002<\/p>\n\n\n\n

\ud83d\udc4dWriting Principles for Task-Tuned Prompt Engineering \u7b46\u8a18<\/a><\/h3>\n\n\n\n

\u9019\u5834\u7531 Anthropic<\/a> \u7684\u7814\u7a76\u54e1 Karina Nguyen \u5e36\u4f86\u7684 Prompting \u6f14\u8b1b\uff0c\u6709\u9304\u5f71\u8ddf\u6295\u5f71\u7247\u3002
\u4e00\u5c0f\u6642\u7684\u6f14\u8b1b\uff0c\u4f46\u5167\u5bb9\u5176\u5be6\u883b\u591a\u7684\uff0c\u9700\u8981\u82b1\u9ede\u6642\u9593\u624d\u80fd\u5438\u6536\uff0c\u6211\u6709\u5beb\u4e86\u7b46\u8a18\u6574\u7406<\/a>\uff0c\u4e0d\u4e00\u5b9a\u8981\u770b\u9304\u5f71\u3002<\/p>\n\n\n\n

* \u958b\u5834\u8b1b\u4e86\u70ba\u4f55 Prompting \u5f88\u96e3\uff0c\u4ee5\u53ca\u4e00\u4e9b Principles\u3002
* \u4e00\u500b\u6848\u4f8b\u662f\u8b1b\u8005\u505a\u7684\u63a8\u85a6\u7cfb\u7d71\uff0c\u6839\u64da\u7528\u6236\u8f38\u5165\u63a8\u85a6\u9069\u5408\u7684\u8863\u670d\uff0c\u7d66 Criteria \u548c\u4f7f\u7528 CoT\u3002
* \u89e3\u8aaa\u4ed6\u5011\u7684 Paper: Question Decomposition Improves the Faithfulness of Model-Generated Reasoning<\/a> \u900f\u904e\u5206\u89e3\u6210\u5b50\u554f\u984c\uff0c\u53ef\u4ee5\u63d0\u6607\u6a21\u578b\u56de\u7b54\u80fd\u529b\u3002 Anthropic \u7684\u505a\u6cd5\u8ddf\u5176\u4ed6\u4eba\u505a Decomposition \u53c8\u6709\u9ede\u4e0d\u4e00\u6a23\uff0c\u6709\u8208\u8da3\u7684\u8a71\u53ef\u4ee5\u4ed4\u7d30\u7814\u7a76\u5dee\u7570\u3002
* \u89e3\u8aaa\u4ed6\u5011 Blog: Prompt engineering for Claude’s long context window<\/a> \u9019\u500b\u5be6\u9a57\u5f88\u6709\u8da3\uff0c\u53ef\u4ee5\u4e00\u8b80\u3002<\/p>\n\n\n\n\n\n\n\n

\u4ed6\u5011\u9078\u4e86\u4e00\u4efd\u653f\u5e9c\u6587\u4ef6\uff0c\u9996\u5148\u62c6 chunks \u7528 prompt \u7522\u751f\u591a\u9078\u984c\uff0c\u7136\u5f8c\u518d\u7528 (\u591a\u9078\u984c+\u76f8\u95dc\u7684\u90a3\u4e00\u500bchunk+\u5176\u4ed6\u7121\u95dc\u7684\u96a8\u6a5fchunks\u585e\u6eff95k\u9577\u5ea6) \u9019\u6a23\u7d44\u8d77\u4f86\u53bb\u8a55\u4f30\u6a21\u578b\u56de\u7b54\u554f\u984c\u7684\u80fd\u529b\u3002<\/p>\n\n\n\n


\u7d50\u8ad6\u662f: <\/p>\n\n\n\n

1.\u5728 doc Q&A \u4e2d\uff0c\u5c07\u554f\u984c\u653e\u5728\u7d50\u5c3e\u554f\uff0c\u6548\u679c\u6bd4\u958b\u982d\u597d
2. \u76f8\u95dc\u5167\u5bb9\u653e\u5728\u7d50\u5c3e\u6642\uff0c\u6548\u679c\u6bd4\u958b\u982d\u597d
3. \u7528\u5176\u4ed6\u4e0d\u76f8\u95dc\u7684\u96a8\u6a5f few-shots \u6c92\u5e6b\u52a9
4. 5 shots \u6bd4 2 shots \u597d
5. \u7528 scratchpad \u6280\u5de7\u6709\u5e6b\u52a9\uff0c\u9019\u62db\u662f\u6307 prompt \u6307\u793a\u6a21\u578b\u5148\u7d66 2-3 \u500b\u76f8\u95dc\u5167\u5bb9\u518d\u7d66\u7b54\u6848<\/p>\n\n\n\n


\u6700\u5f8c\u9084\u6709\u4e00\u4e9b Tips \u8ddf\u7b56\u7565\u4e5f\u5f88\u4e0d\u932f<\/p>\n\n\n\n

1. \u53ef\u4ee5\u628a\u8a71\u585e\u5230 AI \u5634\u88e1\uff0c\u4e5f\u5c31\u662f\u547c\u53eb API \u6642\uff0c\u9664\u4e86 user \u7684\u8a0a\u606f\uff0c\u4f60\u4e5f\u7d66\u4e86 assistant \u7684\u56de\u61c9\u8a0a\u606f\u7576\u4f5c\u7bc4\u4f8b\u3002
2. \u8b93 assistant \u91cd\u8907\u4f60\u7684\u6307\u793a\u3002\u800c\u4e14\u9019\u500b\u91cd\u8907\u7684\u6307\u793a\uff0c\u4e5f\u662f\u4f60\u585e\u5230\u6a21\u578b\u5634\u88e1\u7684\u3002
3. \u6307\u793a\u6a21\u578b\u80fd\u8aaa I don’t know<\/p>\n\n\n\n

\u8a73\u7d30\u53ef\u4ee5\u770b\u6211\u7684\u7b46\u8a18\u642d\u914d\u6295\u5f71\u7247\u7406\u89e3\u3002
<\/p>\n\n\n\n

\ud83d\udd1dAI Engineering 201 \u6295\u5f71\u7247<\/a><\/h3>\n\n\n\n

\u8b1b\u8005 Charles Frye \u662f The Full Stack<\/a> \u7684\u6df1\u5ea6\u5b78\u7fd2\u8001\u5e2b\uff0c\u4ed6\u5011\u7684 LLM \u8ab2\u7a0b\u5f88\u8b9a\u4e5f\u6709\u91cb\u51fa\u9304\u5f71(\u6211\u7684\u7b46\u8a18\u5728\u9019\u88e1<\/a>)\u3002<\/p>\n\n\n\n

\u56de\u5230\u9019\u4efd\u6295\u5f71\u7247\u5167\u5bb9\u4e5f\u4e0d\u5c11\uff0c\u53ef\u60dc\u6c92\u627e\u5230\u9304\u5f71\u3002\u5167\u5bb9\u6bd4\u8f03\u9032\u968e\u5305\u62ec\uff1a<\/p>\n\n\n\n

Part 1 \u8ac7 Running Inference\uff0c\u9700\u8981\u8655\u7406\u90e8\u7f72\u7684\u6a5f\u5668\u5b78\u7fd2\u5de5\u7a0b\u5e2b\u53ef\u4ee5\u770b\u770b
Part 2 \u8ac7 The Rest of the F*cking Owl \u5167\u5bb9\u662f \u67b6\u69cb\u6a21\u5f0f\u3001\u76e3\u63a7\u8ddf\u8a55\u4f30\u7b49<\/p>\n\n\n\n

\ud83c\udfaf\u5176\u4ed6 AI Engineer Summit Workshop \u8cc7\u6599<\/p>\n\n\n\n

* AI Engineering 101: [\u6295\u5f71\u7247]<\/a>
* Finetuning: [\u9304\u5f71]<\/a> \u4f7f\u7528 gradient.ai\/<\/a> \u4f86\u505a\u5fae\u8abf
* Llamaindex \u7684 Building, Evaluating, and Optimizing your RAG App for Production [\u7a0b\u5f0f\u78bc]<\/a> [\u6295\u5f71\u7247]<\/a>
* LangChain \u7684 Function calling and tool usage with LangChain and OpenAI [\u7a0b\u5f0f\u78bc] <\/a>
<\/p>\n\n\n\n

\ud83d\udc4a\u66f4\u591a Claude \u51fa\u7684 Prompt Engineering \u5b78\u7fd2\u8cc7\u6599<\/h3>\n\n\n\n

\u4e0a\u9762\u90a3\u5834 Prompt Engineering \u6f14\u8b1b\uff0c\u8b93\u6211\u5c0d Anthropic \u7684 Claude \u6a21\u578b\u611f\u5230\u5f88\u6709\u8208\u8da3\u3002
\u82e5\u4f60\u6709\u7533\u8acb\u5b98\u65b9\u7684 Claude API \u7684\u8a71\uff0c\u5728\u7533\u8acb\u901a\u904e\u4e4b\u5f8c\u7684\u90f5\u4ef6\u6709\u7d66\u4e00\u4e9b\u5b78\u7fd2\u8cc7\u6599: <\/p>\n\n\n\n

* Prompt design documentation<\/a> \u9019\u662f\u5b98\u65b9\u6587\u4ef6
* Overview of Anthropic models and prompting \u6295\u5f71\u7247<\/a> \u9019\u670924\u9801
* Prompting Basics Course \u6295\u5f71\u7247<\/a> \u9019\u670962\u9801<\/p>\n\n\n\n

\u4e0d\u50cf OpenAI \u7684 Chat API \u4f7f\u7528 messages \u9663\u5217\u53c3\u6578\uff0cClaude \u53ea\u7528\u4e00\u500b prompt \u5b57\u4e32\u53c3\u6578\u3002\u4f46\u9019\u500b\u5b57\u4e32\u4e00\u5b9a\u662f Human: \u958b\u982d\uff0c\u7136\u5f8c\u5176\u4e2d\u7528 Assistant: \u4ee3\u8868\u6a21\u578b\u56de\u8986\u3002<\/p>\n\n\n\n

—-<\/p>\n\n\n\n


\u5e0c\u671b\u4f60\u6703\u559c\u6b61\u9019\u4efd\u9031\u5831\uff0c\u4e0b\u9031\u5c31\u662f OpenAI DevDay<\/a> \u4e86\uff0c\u64da\u8aaa\u6703\u6709 GPT-4V API \u4ee5\u53ca API \u5927\u964d\u50f9\uff0c\u597d\u671f\u5f85\u554a\u554a\u554a\u3002<\/p>\n\n\n\n

\u2013 ihower<\/p>\n","protected":false},"excerpt":{"rendered":"

Hello! \u4f60\u597d \ud83d\udc4b AI Engineer Summit \u5df2\u7d93\u904e\u53bb\u5feb\u4e00\u500b\u6708\uff0c\u5b98\u65b9\u9664\u4e86\u9678\u7e8c\u91cb\u51fa\u5f8c\u88fd … <\/p>\n

\u95b1\u8b80\u5168\u6587\u3008AI Engineer \u9031\u5831\u00a0\ud83d\ude80 Writing Principles for Task-Tuned Prompt Engineering \u7b46\u8a18 #03\u3009<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":"","jetpack_publicize_message":"","jetpack_publicize_feature_enabled":true,"jetpack_social_post_already_shared":true,"jetpack_social_options":{"image_generator_settings":{"template":"highway","default_image_id":0,"font":"","enabled":false},"version":2}},"categories":[91],"tags":[],"class_list":["post-11832","post","type-post","status-publish","format-standard","hentry","category-aie","entry"],"jetpack_publicize_connections":[],"jetpack_featured_media_url":"","jetpack_shortlink":"https:\/\/wp.me\/p1q6tG-34Q","jetpack_sharing_enabled":true,"jetpack_likes_enabled":true,"_links":{"self":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/11832","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/comments?post=11832"}],"version-history":[{"count":8,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/11832\/revisions"}],"predecessor-version":[{"id":13031,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/posts\/11832\/revisions\/13031"}],"wp:attachment":[{"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/media?parent=11832"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/categories?post=11832"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/ihower.tw\/blog\/wp-json\/wp\/v2\/tags?post=11832"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}