DBATU · Second Year B.Tech Artificial Intelligence & Allied · NEP 2020
Prompt Engineering
25AFAIPC307 · Semester III · 2 credits · Continuous assessment 20 + mid-semester 20 + end-semester 60
Unit 1: Introduction to Prompt Engineering and LLMs
4 hoursFrom NLP to large language models; tokens, embeddings and context windows; why prompt engineering matters and where it is used; GPT, Claude, PaLM, Mistral and others; prompting versus fine-tuning.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- From NLP to large language models
- Tokens, embeddings and context windows
- Why prompt engineering matters
- Popular large language models
- Prompting versus fine-tuning
Unit 2: Prompt Design and Advanced Prompting
4 hoursZero-, one- and few-shot prompting; instructional, delimiting, role-based and Socratic patterns; evaluating prompts for accuracy, relevance, hallucination and safety, and debugging them; chain-of-thought, self-ask, ReAct and multimodal prompting; prompt templates with LangChain and the OpenAI SDK.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Zero-shot, one-shot and few-shot prompting
- Prompt patterns
- Evaluating and debugging prompts
- Chain-of-thought and self-ask prompting
- ReAct and multimodal prompting
- Prompt templates with LangChain and the OpenAI SDK
Unit 3: Prompting for Domain-Specific Tasks
4 hoursPrompts for code generation, conversational agents and dialogue design, legal, educational and medical tasks, data wrangling and analysis, and low-resource languages.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Prompts for coding assistance
- Conversational agents and dialogue design
- Domain prompts: legal, education and medicine
- Data analysis through prompting
- Prompting in low-resource languages
Unit 4: Prompt Safety, Bias and Ethics
4 hoursBias, toxicity and misinformation in outputs; adversarial prompting and jailbreaks; safe prompt design; evaluation with BLEU, ROUGE, BERTScore and human review; function calling, tool use and retrieval-augmented generation.
Examined in the 60-mark end-semester paper, and in the mid-semester exam if it is taught by then
- Bias, toxicity and misinformation in LLM outputs
- Adversarial prompts and jailbreaks
- Safe prompt design
- Evaluating LLM outputs: BLEU, ROUGE and BERTScore
- Function calling, tool use and RAG
Unit contents are summarised from the official DBATU syllabus . AiCanCode.org is not affiliated with Dr. Babasaheb Ambedkar Technological University (DBATU).