LLM Chronicles #6.4: LLM Agents with ReAct (Reason + Act)

In this episode we’ll cover LLM agents, focusing on the core research that helped to improve LLMs’ reasoning while allowing them to interact with the external world via the use of tools. These include Chain of Thought prompting, PAL (Program-aided Language Models) and ReAct (Reason + Act) as used in Langchain and CrewAI agents.
Series website: llm-chronicles...
🖹 Canvas:
- llm-chronicles...
🕤 Timestamps:
00:13 - Table of Contents
01:23 - Chain of Thought Prompting
03:10 - PAL (Program-aided Language Models)
05:14 - ReAct (Reason + Act)
09:22 - Tools, Plugins, Functions, APIs
10:54 - ReAct in Practice (JSON/XML formats, fine-tuned models)
12:05 - Function Calling (OpenAI)
13:08 - Modified ReAct (Browser agents, CodeAct)
14:15 - Summary
14:47 - Limitations & Cyber Security Considerations
References:
- Chain-of-Thought Prompting Elicits Reasoning in Large Language Models, arxiv.org/abs/...
- Large Language Models are Zero-Shot Reasoners, arxiv.org/abs/...
- PAL: Program-aided Language Models, arxiv.org/abs/...
- ReAct: Synergizing Reasoning and Acting in Language Models, arxiv.org/abs/...
- InternLM: github.com/Int...
- Executable Code Actions Elicit Better LLM Agents, arxiv.org/pdf/...
- OpenDevin CodeACT: xwang.dev/blog...