Unleashing the potential of prompt engineering for large language models

Abstract

This comprehensive review delves into the pivotal role of prompt engineering in unleashing the capabilities of Large Language Models (LLMs). The development of Artificial Intelligence (AI), from its inception in the 1950s to the emergence of advanced neural networks and deep learning architectures, has made a breakthrough in LLMs, with models such as GPT-4o and Claude-3, and in Vision-Language Models (VLMs), with models such as CLIP and ALIGN. Prompt engineering is the process of structuring inputs, which has emerged as a crucial technique to maximize the utility and accuracy of these models. This paper explores both foundational and advanced methodologies of prompt engineering, including techniques such as self-consistency, chain-of-thought, and generated knowledge, which significantly enhance model performance. Additionally, it examines the prompt method of VLMs through innovative approaches such as Context Optimization (CoOp), Conditional Context Optimization (CoCoOp), and Multimodal Prompt Learning (MaPLe). Critical to this discussion is the aspect of AI security, particularly adversarial attacks that exploit vulnerabilities in prompt engineering. Strategies to mitigate these risks and enhance model robustness are thoroughly reviewed. The evaluation of prompt methods is also addressed through both subjective and objective metrics, ensuring a robust analysis of their efficacy. This review also reflects the essential role of prompt engineering in advancing AI capabilities, providing a structured framework for future research and application.

v6 - Metadata updated (title, journal ref, DOI). PDF identical to v5 (original submission). Please cite the peer-reviewed Version of Record in "Patterns" (DOI: 10.1016/j.patter.2025.101260)

References in corpus (73)

• Explaining and Harnessing Adversarial Examples
• Survey of Hallucination in Natural Language Generation
• Chain-of-Thought Prompting Elicits Reasoning in Large Language Models
• Character-level Convolutional Networks for Text Classification
• Learning to Prompt for Vision-Language Models
• Llama 2: Open Foundation and Fine-Tuned Chat Models
• PaLM: Scaling Language Modeling with Pathways
• Unsupervised Data Augmentation for Consistency Training
• Sparks of Artificial General Intelligence: Early experiments with GPT-4
• Zero-Shot Text-to-Image Generation
• Large Language Models are Zero-Shot Reasoners
• Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
• A Survey on Large Language Model (LLM) Security and Privacy: The Good, the Bad, and the Ugly
• BadNets: Identifying Vulnerabilities in the Machine Learning Model Supply Chain
• Gemini: A Family of Highly Capable Multimodal Models
• A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT
• A Survey on Multimodal Large Language Models
• Stealing Machine Learning Models via Prediction APIs
• Beyond the Imitation Game: Quantifying and extrapolating the capabilities of language models
• Think you have Solved Question Answering? Try ARC, the AI2 Reasoning Challenge
• Graph of Thoughts: Solving Elaborate Problems with Large Language Models
• Toolformer: Language Models Can Teach Themselves to Use Tools
• CommonsenseQA: A Question Answering Challenge Targeting Commonsense Knowledge
• Solving Quantitative Reasoning Problems with Language Models
• Visual Question Answering: Datasets, Algorithms, and Future Challenges
• The Rise and Potential of Large Language Model Based Agents: A Survey
• A Systematic Survey of Prompt Engineering in Large Language Models: Techniques and Applications
• Universal and Transferable Adversarial Attacks on Aligned Language Models
• Challenges and Applications of Large Language Models
• How Deep Learning Sees the World: A Survey on Adversarial Attacks & Defenses
• I Know What You Trained Last Summer: A Survey on Stealing Machine Learning Models and Defences
• Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks
• A very preliminary analysis of DALL-E 2
• Unsolved Problems in ML Safety
• Prompt Injection attack against LLM-integrated Applications
• Internet-augmented language models through few-shot prompting for open-domain question answering
• A Systematic Survey of Prompt Engineering on Vision-Language Foundation Models
• Identifying and Extracting Rare Disease Phenotypes with Large Language Models
• Evaluating Human-Language Model Interaction
• ExpertPrompting: Instructing Large Language Models to be Distinguished Experts
• Large Language Model Guided Tree-of-Thought
• WebShop: Towards Scalable Real-World Web Interaction with Grounded Language Agents
• Agent AI: Surveying the Horizons of Multimodal Interaction
• The AI Revolution: Opportunities and Challenges for the Finance Sector
• Survey of Vulnerabilities in Large Language Models Revealed by Adversarial Attacks
• LLM Multi-Agent Systems: Challenges and Open Problems
• Teaching Algorithmic Reasoning via In-context Learning
• Solving math word problems with process- and outcome-based feedback
• Evaluation of OpenAI o1: Opportunities and Challenges of AGI
• Scaling Transformer to 1M tokens and beyond with RMT
• Meta-Prompting: Enhancing Language Models with Task-Agnostic Scaffolding
• Adversarial Demonstration Attacks on Large Language Models
• Risk or Chance? Large Language Models and Reproducibility in HCI Research
• Bring Your Own Data! Self-Supervised Evaluation for Large Language Models
• Analyzing Chain-of-Thought Prompting in Large Language Models via Gradient-based Feature Attributions
• Large Language Models Perform Diagnostic Reasoning
• Walking Down the Memory Maze: Beyond Context Limit through Interactive Reading
• Teaching Arithmetic to Small Transformers
• Exploring Prompt Engineering Practices in the Enterprise
• Securing Large Language Models: Threats, Vulnerabilities and Responsible Practices
• If our aim is to build morality into an artificial agent, how might we begin to go about doing so?
• Jailbreak Attacks and Defenses Against Large Language Models: A Survey
• Prompt Stealing Attacks Against Large Language Models
• Investigating Prompting Techniques for Zero- and Few-Shot Visual Question Answering
• AQuA: A Benchmarking Tool for Label Quality Assessment
• Post Hoc Explanations of Language Models Can Improve Language Models
• Guided scenarios with simulated expert personae: a remarkable strategy to perform cognitive work
• Exploring Vulnerabilities and Protections in Large Language Models: A Survey
• PoisonBench: Assessing Large Language Model Vulnerability to Data Poisoning
• Chain-of-Thought Reasoning is a Policy Improvement Operator
• TreePrompt: Learning to Compose Tree Prompts for Explainable Visual Grounding
• Decoding Decoded: Understanding Hyperparameter Effects in Open-Ended Text Generation
• ASETF: A Novel Method for Jailbreak Attack on LLMs through Translate Suffix Embeddings

Cited by in corpus (6)

• RoBERTa-BiLSTM: A Context-Aware Hybrid Model for Sentiment Analysis
• Metamorphic Testing of Large Language Models for Natural Language Processing
• Computational Hermeneutics: Evaluating generative AI as a cultural technology
• Textual interpretation of transient image classifications from large language models
• Improving Lesion Segmentation in Medical Images by Global and Regional Feature Compensation
• MetaSpace: Metamorphic Testing for Spatial Cognition in Embodied Agents