Introduction

1.8 Agents

1. What is an agent?
   An LLM that can act autonomously in the world by calling other programs (search engine, calendar, calculator, database, supplier API) through actions like SEARCH() that it can generate in its output.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 1, §1.8, p. 24
2. What is the technical difference between a conversational LM and an agent?
   Only that actions in the world are added to the set of possible tokens to generate; it is still token prediction.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 1, §1.8, p. 24
3. How does the ReAct framework work?
   The LLM is prompted to interleave reasoning traces with actions: it loops over Reason → Act (call a tool) → Observe (read the result), with the whole history kept as textual context, until the task is solved (e.g. Finish[answer]).

   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 1, §1.8, pp. 24–25