Summarization
Summarization is one of the most common LLM tasks. You provide a long text and ask the model to condense it.
You can control the output format — bullet points, one sentence, a paragraph — by stating it in your prompt.
Prompt
Summarize the following article in 3 bullet points:
The James Webb Space Telescope (JWST) has captured its most
detailed images yet of the Crab Nebula, a supernova remnant
located about 6,500 light-years from Earth. The new infrared
observations reveal previously unseen structures in the
nebula's interior, including filaments of dust that glow in
wavelengths invisible to earlier telescopes. Scientists say
the data will help them understand how dying stars seed the
universe with heavy elements. The Crab Nebula was created by
a supernova explosion observed by Chinese astronomers in
1054 AD.
The model extracts the key points and formats them exactly as requested.
Response
• JWST captured the most detailed infrared images yet of
the Crab Nebula, revealing previously unseen dust
filaments and internal structures.
• The observations will help scientists understand how
supernova remnants distribute heavy elements across
the universe.
• The Crab Nebula was formed by a supernova explosion
first recorded by Chinese astronomers in 1054 AD.
You can also ask for different summary styles. Here we ask for a single-sentence summary.
Prompt
Summarize the same article in one sentence.
The model adapts its output length to match your request.
Response
JWST's new infrared images of the Crab Nebula reveal
previously hidden structures that will help scientists
understand how dying stars enrich the universe with
heavy elements.
Key takeaway: Be specific about the format you want. "Summarize in 3 bullet points" gives very different results from "Summarize in one paragraph." The more precise your instruction, the better the output.
Tips:
- Specify length: "in 50 words", "in 3 bullets"
- Specify audience: "for a 5th grader", "for experts"
- Specify focus: "focusing on the business impact"