利用 GenAI_ML 进行 OCP 错误分类：优化测试错误分类周期.pdf

1、OCP Error Categorization with GenAI/ML:Optimizing the test error categorization cycleOCP Error Categorization with GenAI:Optimizing the test error categorization cycleArushi Sharma,MetaGustavo Castellanos,MetaRicha Mishra,MetaSYSTEMS MANAGEMENTHardware failures during the NPI phaseis an inevitable p

2、art of the design,test,and validation process.Thousands of logs get generated with each new platform testing,after comprehensive testing.Provides insights on key areaswith high failure rates.Introduction-Error Categorization during NPIManual error mapping ofnew error signatures to existing or releva

3、nt error types Time consumingRequires domain expertiseDifficulty in identifying new error typesLack of scalability and flexibility for new platformsProblem StatementTargetTrained on vast amounts of dataCan be fine-tuned for specific tasks like error categorization Automated and optimized process Adv

4、anced techniques for efficientcategorization Scalable and adaptable solution Improved EfficiencyLLMs for Error Categorization1.No training/No Fine tuning a.Zero shot prompting b.Chain of Thought(CoT)c.Chain of Draft(CoD)1.Semantic Searchb.Cosine/Dot Product/Euclidean Similarity(FAISS)c.Dragon-RoBERT

5、a/DRAMA Dense retrievers1.Training/Fine Tuning b.Fine tuning Llama with LORA(Low-Rank Adaptation of LLMs)using QLora/Other Open Source ModelsIntroduction to Gen AI ApproachesOption 1-LLM PromptingZero Shot PromptingCategories documentMetagen LLMin-contextpromptCategoryLog/Error to categorizeCategori

6、es documentMetagen LLMin-contextpromptCategoryLog/Error to categorizechallengelogicCoT AgentChain of Thought Prompting(CoT)Chain of Draft Prompting(CoD)Metagen LLMLLM Prompting in ActionLLMCategoryError SignatureOption 2-Semantic Search/Dense RetrieversCategories documentVector DB CategoriesEmbeddin