Скачать или смотреть LLMs | LLMs and Tools: Function Calling | Lec 18.2

tl;dr: This lecture covers the methodologies for enabling LLMs to call and utilize external APIs, discussing techniques to train models for selecting and using functions as needed, thereby expanding their practical capabilities in interacting with the external digital environment.

🎓 Lecturer: Dinesh Raghu [https://research.ibm.com/people/dines...]
🔗 Get the Slides Here: http://lcs2.in/llm2401
📚 Suggested Readings:
[Gorilla: Large Language Model Connected with Massive APIs](https://arxiv.org/pdf/2305.15334)
[ToolAlpaca: Generalized Tool Learning for Language Models with 3000 Simulated Cases](https://arxiv.org/pdf/2306.05301)
[ToolLLM: Facilitating Large Language Models to Master 16000+ Real-world APIs](https://arxiv.org/pdf/2307.16789)
[APIGen: Automated Pipeline for Generating Verifiable and Diverse Function-Calling Datasets](https://arxiv.org/pdf/2406.18518)
[Granite-Function Calling Model: Introducing Function Calling Abilities via Multi-task Learning of Granular Tasks](https://arxiv.org/pdf/2407.00121)

In this lecture, we delve into the techniques for teaching Large Language Models (LLMs) to interface with APIs and effectively call appropriate functions. By enabling function calling capabilities, LLMs can perform tasks such as data retrieval, computations, and interacting with real-world applications. We’ll explore the architecture and strategies behind equipping models to understand when and how to use specific APIs, enhancing their utility in practical, dynamic scenarios.