Building with agents 0/15

Building with agents

Prompts and context

What goes into one model call: the system prompt, the conversation, retrieved text and tool results, all inside one context window. The model can use only what is in that window, so choosing it and writing a checkable task decide the answer.

Reading

WhiteBox lessons3

Papers and essays2

  1. Chain-of-Thought Prompting Elicits Reasoning in Large Language Models(opens in a new tab)

    Wei et al. 2022 NeurIPS 2022 40 min

    Chain-of-thought prompting: few-shot examples that show intermediate reasoning steps lead large models to write their own steps, raising accuracy on arithmetic, commonsense and symbolic reasoning tasks.

  2. Lost in the Middle: How Language Models Use Long Contexts(opens in a new tab)

    Liu et al. 2024 Transactions of the ACL 40 min

    Language models use relevant information at the start or end of a long context much better than information in the middle, and this holds even for models built for long contexts.

Videos1

Building with LLMs Various speakers 1 video