Lesson 2 · 7 topics · 20 min · intermediate

Few-Shot Examples and Structured Outputs

You will be able to

Explain in-context learning, zero-shot and few-shot prompting, and what AI Engineering reports about how much examples help stronger models.
Choose and order few-shot examples: cover the labels, keep the format compact, end on a marker, and test the order.
Ask for a structured output precisely: keys, types, allowed labels and an example.
Name the ways structured outputs break (truncation, extra prose, wrong types, invented labels, missing fields) and tell parse failures from schema failures.
Compare prompting, post-processing, test time compute, constrained sampling and finetuning, and say what each can guarantee.
Validate outputs against a schema, repair syntax, retry content failures, and measure parse, valid and correct rates on a fixed set.

Topics

Your review

Sign in to keep your quick-check answers. The ones you miss come back for review after 1, 3 and 7 days, until they stick.

Sources

AI Engineering · Chip Huyen, 2025
Chip Huyen. O'Reilly. Used: Ch. 5 "Prompt Engineering": "In-Context Learning: Zero-Shot and Few-Shot"; "Write Clear and Explicit Instructions" > "Explain, without ambiguity, what you want the model to do", "Provide examples" (Table 5-2), "Specify the output format" (Table 5-3). Ch. 2 "Understanding Foundation Models": "Structured Outputs" > "Prompting", "Post-processing", "Constrained sampling", "Finetuning"; "Test Time Compute".
Hands-On Large Language Models · Jay Alammar & Maarten Grootendorst, 2024
Jay Alammar & Maarten Grootendorst. O'Reilly. Used: Ch. 6 "Prompt Engineering": "In-Context Learning: Providing Examples"; "Output Verification" > "Providing Examples", "Grammar: Constrained Sampling".
Calibrate Before Use: Improving Few-Shot Performance of Language Models · Tony Z. Zhao et al., 2021
Tony Z. Zhao, Eric Wallace, Shi Feng, Dan Klein, Sameer Singh. Used: Abstract, fetched 2026-10-21. Open source
Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity · Yao Lu et al., 2021
Yao Lu, Max Bartolo, Alastair Moore, Sebastian Riedel, Pontus Stenetorp. Used: Abstract, fetched 2026-10-21. Open source

You will be able to

Explain in-context learning, zero-shot and few-shot prompting, and what AI Engineering reports about how much examples help stronger models.

Choose and order few-shot examples: cover the labels, keep the format compact, end on a marker, and test the order.

Ask for a structured output precisely: keys, types, allowed labels and an example.

Name the ways structured outputs break (truncation, extra prose, wrong types, invented labels, missing fields) and tell parse failures from schema failures.

Compare prompting, post-processing, test time compute, constrained sampling and finetuning, and say what each can guarantee.

Validate outputs against a schema, repair syntax, retry content failures, and measure parse, valid and correct rates on a fixed set.

Sources

AI Engineering · Chip Huyen, 2025

Chip Huyen. O'Reilly. Used: Ch. 5 "Prompt Engineering": "In-Context Learning: Zero-Shot and Few-Shot"; "Write Clear and Explicit Instructions" > "Explain, without ambiguity, what you want the model to do", "Provide examples" (Table 5-2), "Specify the output format" (Table 5-3). Ch. 2 "Understanding Foundation Models": "Structured Outputs" > "Prompting", "Post-processing", "Constrained sampling", "Finetuning"; "Test Time Compute".

Hands-On Large Language Models · Jay Alammar & Maarten Grootendorst, 2024

Jay Alammar & Maarten Grootendorst. O'Reilly. Used: Ch. 6 "Prompt Engineering": "In-Context Learning: Providing Examples"; "Output Verification" > "Providing Examples", "Grammar: Constrained Sampling".

Calibrate Before Use: Improving Few-Shot Performance of Language Models · Tony Z. Zhao et al., 2021

Tony Z. Zhao, Eric Wallace, Shi Feng, Dan Klein, Sameer Singh. Used: Abstract, fetched 2026-10-21. Open source

Fantastically Ordered Prompts and Where to Find Them: Overcoming Few-Shot Prompt Order Sensitivity · Yao Lu et al., 2021

Yao Lu, Max Bartolo, Alastair Moore, Sebastian Riedel, Pontus Stenetorp. Used: Abstract, fetched 2026-10-21. Open source