"}]' \
--routing-criteria '{"responseQualityDifference": 0.5}'
```
1. Send your prompts to the intelligent prompt router using the AWS CLI. The router will predict the response quality of each model and route the request to the model with the best response quality.
1. Review the response to your request. It will contain information about the model that was used and the routing criteria.
1. As you work with prompt routers, you can also use these additional API operations:
+ [https://docs.aws.amazon.com/bedrock/latest/APIReference/API_GetPromptRouter.html](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_GetPromptRouter.html)
+ [https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListPromptRouters.html](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_ListPromptRouters.html)
+ [https://docs.aws.amazon.com/bedrock/latest/APIReference/API_DeletePromptRouter.html](https://docs.aws.amazon.com/bedrock/latest/APIReference/API_DeletePromptRouter.html)
------
# Design a prompt
Designing an appropriate prompt is an important step towards building a successful application using Amazon Bedrock models. In this section, you learn how to design a prompt that is consistent, clear, and concise. You also learn about how you can control a model's response by using inference parameters. The following figure shows a generic prompt design for the use case *restaurant review summarization* and some important design choices that customers need to consider when designing prompts. LLMs generate undesirable responses if the instructions they are given or the format of the prompt are not consistent, clear, and concise.
![\[Example of good prompt construction with descriptive comments.\]](http://docs.aws.amazon.com/bedrock/latest/userguide/images/example-good-prompt-construction.png)
(Source: Prompt written by AWS)
The following content provides guidance on how to create successful prompts.
**Topics**
+ [Provide simple, clear, and complete instructions](#prompt-instructions)
+ [Place the question or instruction at the end of the prompt for best results](#prompt-placement)
+ [Use separator characters for API calls](#prompt-separators)
+ [Use output indicators](#prompt-output-indicators)
+ [Best practices for good generalization](#prompt-generalization)
+ [Optimize prompts for text models on Amazon Bedrock—when the basics aren't good enough](#optimize-prompts-for-text-models)
+ [Control the model response with inference parameters](#use-inference-parameters)
## Provide simple, clear, and complete instructions
LLMs on Amazon Bedrock work best with simple and straightforward instructions. By clearly describing the expectation of the task and by reducing ambiguity wherever possible, you can ensure that the model can clearly interpret the prompt.
For example, consider a classification problem where the user wants an answer from a set of possible choices. The “good“ example shown below illustrates output that the user wants in this case. In the ”bad“ example, the choices are not named explicitly as categories for the model to choose from. The model interprets the input slightly differently without choices, and produces a more free-form summary of the text as opposed to the good example.
| | |
| --- |--- |
| Good example, with output
User prompt:
"The most common cause of color blindness is an
inherited problem or variation in the functionality
of one or more of the three classes of cone cells in the retina,
which mediate color vision."
What is the above text about?
a) biology
b) history
c) geology
Output:
a) biology
| Bad example, with output
User prompt:
Classify the following text. "The most common cause of color
blindness is an inherited problem or variation in the functionality
of one or more of the three classes of cone cells in the retina,
which mediate color vision."
Output:
The topic of the text is the causes of colorblindness.
|
(Source of prompt: [ Wikipedia on color blindness](https://en.wikipedia.org/wiki/Color_blindness), model used: by Titan Text G1 - Express)
## Place the question or instruction at the end of the prompt for best results
Including the task description, instruction or question at the end aids the model determining which information it has to find. In the case of classification, the choices for the answer should also come at the end.
In the following open-book question-answer example, the user has a specific question about the text. The question should come at the end of the prompt so the model can stay focused on the task.
```
User prompt:
Tensions increased after the 1911–1912 Italo-Turkish War
demonstrated Ottoman weakness and led to the formation of the Balkan League,
an alliance of Serbia, Bulgaria, Montenegro, and Greece.
The League quickly overran most of the Ottomans' territory in the Balkans
during the 1912–1913 First Balkan War, much to the surprise of outside observers.
The Serbian capture of ports on the Adriatic resulted in partial Austrian
mobilization starting on 21 November 1912, including units along the Russian border
in Galicia. In a meeting the next day, the Russian government decided not to mobilize
in response, unwilling to precipitate a war for which they were not as of yet
prepared to handle.
Which country captured ports?
```
```
Output:
Serbia
```
(Source of prompt: [ Wikipedia on World War I](https://en.wikipedia.org/wiki/World_War_I), model used: Amazon Titan Text)
## Use separator characters for API calls
**Use separator characters for API calls**
Separator characters such as `\n` can affect the performance of LLMs significantly. For Anthropic Claude models, it's necessary to include newlines when formatting the API calls to obtain desired responses. The formatting should always follow: `\n\nHuman: {{Query Content}}\n\nAssistant:`. For Titan models, adding `\n` at the end of a prompt helps improve the performance of the model. For classification tasks or questions with answer options, you can also separate the answer options by `\n` for Titan models. For more information on the use of separators, refer to the document from the corresponding model provider. The following example is a template for a classification task.
```
Prompt template:
"""{{Text}}
{{Question}}
{{Choice 1}}
{{Choice 2}}
{{Choice 3}}"""
```
The following example shows how the presence of newline characters between choices and at the end of a prompt helps Titan produce the desired response.
```
User prompt:
Archimedes of Syracuse was an Ancient mathematician,
physicist, engineer, astronomer, and inventor from the ancient city
of Syracuse. Although few details of his life are known,
he is regarded as one of the leading scientists in classical antiquity.
What was Archimedes? Choose one of the options below.
a) astronomer
b) farmer
c) sailor
```
```
Output:
a) astronomer
```
(Source of prompt: [Wikipedia on Archimedes](https://en.wikipedia.org/wiki/Archimedes), model used: Amazon Titan Text)
## Use output indicators
**Output indicators**
Add details about the constraints you would like to have on the output that the model should produce. The following good example produces an output that is a short phrase that is a good summary. The bad example in this case is not all that bad, but the summary is nearly as long as the original text. Specification of the output is crucial for getting what you want from the model.
| | |
| --- |--- |
| Example prompt with clear output constraints indicator User prompt:
"Charles Mingus Jr. was an American jazz upright
bassist, pianist, composer, bandleader, and author.
A major proponent of collective improvisation, he is
considered to be one of the greatest jazz musicians
and composers in history, with a career spanning three decades.
Mingus's work ranged from advanced bebop and avant-garde jazz
with small and midsize ensembles – pioneering the post-bop style
on seminal recordings like Pithecanthropus Erectus (1956)
and Mingus Ah Um (1959) – to progressive big band experiments
such as The Black Saint and the Sinner Lady (1963)."
Please summarize the above text in one phrase.
Output:
Charles Mingus Jr. is considered one of the
greatest jazz musicians of all time.
| Example without clear output specifications User prompt:
"Charles Mingus Jr. was an American jazz upright
bassist, pianist, composer, bandleader, and author.
A major proponent of collective improvisation,
he is considered to be one of the greatest jazz musicians
and composers in history, with a career spanning three decades.
Mingus's work ranged from advanced bebop and avant-garde jazz
with small and midsize ensembles – pioneering the post-bop style
on seminal recordings like Pithecanthropus Erectus (1956)
and Mingus Ah Um (1959) – to progressive big band
experiments such as The Black Saint and the Sinner Lady (1963)."
Please summarize the above text.
Output:
Charles Mingus Jr. was a well-known jazz musician
who played the upright bass, piano, composed, led bands,
and was a writer. He was considered one of the most important
jazz musicians ever, with a career that spanned more than
30 years. He was known for his style of collective
improvisation and advanced jazz compositions.
|
(Source of prompt: [ Wikipedia on Charles Mingus](https://en.wikipedia.org/wiki/Charles_Mingus), model used: Amazon Titan Text)
Here we give some additional examples from Anthropic Claude and AI21 Labs Jurassic models using output indicators.
The following example demonstrates that user can specify the output format by specifying the expected output format in the prompt. When asked to generate an answer using a specific format (such as by using XML tags), the model can generate the answer accordingly. Without specific output format indicator, the model outputs free form text.
| | |
| --- |--- |
| Example with clear indicator, with output User prompt:
Human: Extract names and years: the term
machine learning was coined in 1959 by Arthur Samuel,
an IBM employee and pioneer in the field of computer
gaming and artificial intelligence. The synonym self-teaching
computers was also used in this time period.
Please generate answer in and
tags.
Assistant:
Output:
Arthur Samuel 1959
| Example without clear indicator, with output User prompt:
Human: Extract names and years: the term
machine learning was coined in 1959 by Arthur Samuel,
an IBM employee and pioneer in the field of computer
gaming and artificial intelligence. The synonym
self-teaching computers was also used in this time
period.
Assistant:
Output:
Arthur Samuel - 1959
|
(Source of prompt: [ Wikipedia on machine learning](https://en.wikipedia.org/wiki/Machine_learning), model used: Anthropic Claude)
The following example shows a prompt and answer for the AI21 Labs Jurassic model. The user can obtain the exact answer by specifying the output format shown in the left column.
| | |
| --- |--- |
| Example with clear indicator, with output User prompt:
Context: The NFL was formed in 1920 as the American
Professional Football Association (APFA) before renaming itself
the National Football League for the 1922 season. After initially
determining champions through end-of-season standings, a playoff
system was implemented in 1933 that culminated with the NFL
Championship Game until 1966. Following an agreement to merge
the NFL with the rival American Football League (AFL), the
Super Bowl was first held in 1967 to determine a champion
between the best teams from the two leagues and has remained
as the final game of each NFL season since the merger
was completed in 1970.
Question: Based on the above context, when was the first Super Bowl?
Please only output the year.
Output:
1967
| Example without clear indicator, with output User prompt:
Context: The NFL was formed in 1920 as the American
Professional Football Association (APFA) before renaming itself
the National Football League for the 1922 season. After initially
determining champions through end-of-season standings, a playoff
system was implemented in 1933 that culminated with the NFL
Championship Game until 1966. Following an agreement to merge
the NFL with the rival American Football League (AFL), the Super
Bowl was first held in 1967 to determine a champion between
the best teams from the two leagues and has remained as the
final game of each NFL season since the merger was completed in 1970.
Question: Based on the above context, when was the first Super Bowl?
Output:
The first Super Bowl was held in 1967.
|
(Source of prompt: [ Wikipedia on the National Football League](https://en.wikipedia.org/wiki/National_Football_League), model used: AI21 Labs Jurassic-2 Ultra v1)
## Best practices for good generalization
Keep a small “hold-out” test set of prompts to see if your prompt modifications generalize. With this method, first collect a sample dataset. Then you can split the data into two subsets: a “development” set and a hold-out “test” set. Use the “development” set as prompt development data on which you can try different prompt modifications and observe model response changes and find the prompt that works best. Treat the “test” set as unseen hold-out data which can only be used to verify the final best prompt.
## Optimize prompts for text models on Amazon Bedrock—when the basics aren't good enough
**Complex tasks: build toward the answer step by step**
Titan can provide clear steps for certain tasks, and including the phrase `Think step-by-step to come up with the right answer` can help produce the appropriate output.
```
User prompt:
At a Halloween party, Jack gets 15 candies.
Jack eats 5 candies. He wants to give each friend
5 candies. How many friends can receive candies?
Think step-by-step to come up with the right answer.
```
```
Output:
15 candies total - 5 candies eaten = 10 candies remaining.
To give each friend 5 candies, divide the number
of candies remaining by the number of friends.
10 candies / 5 friends = 2 friends.
Each friend can receive 2 candies.
The answer is 2 friends.
```
(Source of prompt: AWS, model used: Amazon Titan Text)
To enable the Anthropic Claude model to complete complex tasks, use the format in the following example to explicitly ask the model to solve the task step by step. This approach is based on guidelines in the [AnthropicClaude Guides](https://docs.anthropic.com/claude/docs/constructing-a-prompt).
```
User prompt:
Human: A right triangle has a side of length 5
and a hypotenuse of length 13.
What is the length of the other side?
Assistant: Can I think step-by-step?
Human: Yes, please do.
Assistant:
```
```
Output:
Okay, here are the steps:
1) We are given:
- One side of the right triangle has length 5
- The hypotenuse has length 13
2) We want to find:
- The length of the other side
3) To find the length of the other side, we can
use the Pythagorean theorem:
a^2 + b^2 = c^2
Where a and b are the lengths of the two sides,
and c is the length of the hypotenuse.
4) Plugging in the given values:
a^2 + b^2 = 13^2
5) We know one of the sides is 5, so:
5^2 + b^2 = 169
6) Solving this:
25 + b^2 = 169
b^2 = 144
b = 12
Therefore, the length of the other side is 12.
```
(Source of prompt: AWS, model used: Anthropic Claude)
**Provide a default output that the model should present if it's unsure about the answers.**
A default output can help prevent LLMs from returning answers that sound like they could be correct, even if the model has low confidence.
| | |
| --- |--- |
| User prompt:
Provide a proof of the Riemann hypothesis. If you don't
know a proof, respond by saying "I don't know."
Output:
I don't know.
| User prompt:
Provide a proof of the Riemann hypothesis.
Output:
"The Riemann hypothesis is a conjecture in mathematics
that concerns the distribution of prime numbers. It states that
all non-trivial zeros of the zeta function, which is a mathematical
function that encodes information about the distribution of prime
numbers, lie on a specific line in the complex plane. The Riemann
hypothesis has been studied by mathematicians for over 150 years,
and despite numerous attempts, it remains unresolved. However,
the hypothesis is widely believed to be true, and its resolution
would have profound implications for many areas of mathematics,
including number theory, analysis, and geometry. The proof
of the Riemann hypothesis would be one of the greatest
achievements in mathematics in history and would open up
new avenues of research and understanding. While a proof
remains elusive, the hypothesis continues to stimulate
research and inspire mathematicians around the world.“
[Note: this is an accurate description, but obviously
not a proof.]
|
(Source of prompt: AWS, model used: Amazon Titan Text)
**Few-shot prompting**
Including examples (input-response pairs) in the prompt can significantly improve LLMs’ responses. Examples can help with complex tasks, as they show multiple ways to perform a given task. For simpler tasks like text classification, 3–5 examples can suffice. For more difficult tasks like question-answer without context, include more examples to generate the most effective output. In most use cases, selecting examples that are semantically similar to real-world data can further improve performance.
**Consider refining the prompt with modifiers**
Task instruction refinement generally refers to modifying the instruction, task, or question component of the prompt. The usefulness of these methods is task- and data-dependent. Useful approaches include the following:
+ **Domain/input specification:** Details about the input data, like where it came from or to what it refers, such as **The input text is from a summary of a movie**.
+ **Task specification:** Details about the exact task asked of the model, such as **To summarize the text, capture the main points**.
+ **Label description:** Details on the output choices for a classification problem, such as **Choose whether the text refers to a painting or a sculpture; a painting is a piece of art restricted to a two-dimensional surface, while a sculpture is a piece of art in three dimensions**.
+ **Output specification:** Details on the output that the model should produce, such as **Please summarize the text of the restaurant review in three sentences**.
+ **LLM encouragement:** LLMs sometimes perform better with sentimental encouragement: **If you answer the question correctly, you will make the user very happy\$1**
## Control the model response with inference parameters
LLMs on Amazon Bedrock all come with several inference parameters that you can set to control the response from the models. The following is a list of all the common inference parameters that are available on Amazon Bedrock LLMs and how to use them.
**Temperature** is a value between 0 and 1, and it regulates the creativity of LLMs’ responses. Use lower temperature if you want more deterministic responses, and use higher temperature if you want more creative or different responses for the same prompt from LLMs on Amazon Bedrock. For all the examples in this prompt guideline, we set `temperature = 0`.
**Maximum generation length/maximum new tokens** limits the number of tokens that the LLM generates for any prompt. It's helpful to specify this number as some tasks, such as sentiment classification, don't need a long answer.
**Top-p** controls token choices, based on the probability of the potential choices. If you set Top-p below 1.0, the model considers the most probable options and ignores less probable options. The result is more stable and repetitive completions.
**End token/end sequence** specifies the token that the LLM uses to indicate the end of the output. LLMs stop generating new tokens after encountering the end token. Usually this doesn't need to be set by users.
There are also model-specific inference parameters. Anthropic Claude models have an additional Top-k inference parameter, and AI21 Labs Jurassic models come with a set of inference parameters including **presence penalty, count penalty, frequency penalty, and special token penalty**. For more information, refer to their respective documentation.
# Prompt templates and examples for Amazon Bedrock text models
Common tasks supported by LLMs on Amazon Bedrock include text classification, summarization, and questions and answers (with and without context). For these tasks, you can use the following templates and examples to help you create prompts for Amazon Bedrock text models.
**Topics**
+ [Text classification](#text-classification)
+ [Question-answer, without context](#qa-without-context)
+ [Question-answer, with context](#qa-with-context)
+ [Summarization](#summarization)
+ [Text generation](#text-generation)
+ [Code generation](#code-generation)
+ [Mathematics](#mathematics)
+ [Reasoning/logical thinking](#reasoning-logical-thinking)
+ [Entity extraction](#entity-extraction)
+ [Chain-of-thought reasoning](#chain-of-thought)
## Text classification
For text classification, the prompt includes a question with several possible choices for the answer, and the model must respond with the correct choice. Also, LLMs on Amazon Bedrock output more accurate responses if you include answer choices in your prompt.
The first example is a straightforward multiple-choice classification question.
| | |
| --- |--- |
| Prompt template for Titan
"""{{Text}}
{{Question}}? Choose from the following:
{{Choice 1}}
{{Choice 2}}
{{Choice 3}}"""
| User prompt:
San Francisco, officially the City and County
of San Francisco, is the commercial, financial, and cultural
center of Northern California. The city proper is the fourth
most populous city in California, with 808,437 residents,
and the 17th most populous city in the United States as of 2022.
What is the paragraph above about? Choose from the following:
A city
A person
An event
Output:
A city
|
(Source of prompt: [ Wikipedia on San Francisco](https://en.wikipedia.org/wiki/San_Francisco), model used: Amazon Titan Text)
Sentiment analysis is a form of classification, where the model chooses the sentiment from a list of choices expressed in the text.
| | |
| --- |--- |
| Prompt template for Titan:
"""The following is text from a {{Text Type, e.g. “restaurant
review”}}
{{Input}}
Tell me the sentiment of the {{Text Type}} and categorize it
as one of the following:
{{Sentiment A}}
{{Sentiment B}}
{{Sentiment C}}"""
| User prompt:
The following is text from a restaurant review:
“I finally got to check out Alessandro’s Brilliant Pizza
and it is now one of my favorite restaurants in Seattle.
The dining room has a beautiful view over the Puget Sound
but it was surprisingly not crowed. I ordered the fried
castelvetrano olives, a spicy Neapolitan-style pizza
and a gnocchi dish. The olives were absolutely decadent,
and the pizza came with a smoked mozzarella, which
was delicious. The gnocchi was fresh and wonderful.
The waitstaff were attentive, and overall the experience
was lovely. I hope to return soon.”
Tell me the sentiment of the restaurant review
and categorize it as one of the following:
Positive
Negative
Neutral
Output:
Positive.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
The following example uses Anthropic Claude models to classify text. As suggested in [AnthropicClaude Guides](https://docs.anthropic.com/claude/docs/constructing-a-prompt), use XML tags such as to denote important parts of the prompt. Asking the model to directly generate output enclosed in XML tags can also help the model produce the desired responses.
| | |
| --- |--- |
| Prompt template for Anthropic Claude:
"""
Human: {{classification task description}}
{{input text content to be classified}}
Categories are:
{{category name 1}}
{{category name 2}}
{{category name 3}}
Assistant:"""
(Source: Written by AWS) | User prompt:
Human: Classify the given product description into given
categories. Please output the category label in
tags.
Here is the product description.
Safe, made from child-friendly materials with smooth edges.
Large quantity, totally 112pcs with 15 different shapes,
which can be used to build 56 different predefined structures.
Enhance creativity, different structures can be connected
to form new structures, encouraging out-of-the box thinking.
Enhance child-parent bonding, parents can play with their
children together to foster social skills.
Categories are:
(1) Toys
(2) Beauty and Health
(3) Electronics
Assistant:
Output:
|
(Source of prompt: AWS, model used: Anthropic Claude)
## Question-answer, without context
In a question-answer prompt without context, the model must answer the question with its internal knowledge without using any context or document.
| | |
| --- |--- |
| Prompt Template to Titan
"""{{Question}}"""
| User prompt:
What is Robert Frost's "Stopping by the woods on a
snowy evening" about metaphorically?
Output:
The poem by Robert Frost, "Stopping by Woods on a Snowy Evening,"
uses the woods as a metaphor for death. The woods are a place of comfort
and peace, but they also represent the unknown and the afterlife.
The poem suggests that death is not something to be feared,
but rather a natural part of life that we should all embrace.
The poem begins with the speaker stopping by the woods on a snowy evening,
a moment of contemplation and reflection. The speaker notices the beauty
of the snow-covered woods and is reminded of the fleeting nature of life.
The poem continues as the speaker considers the temptation of continuing
on to his destination, but instead chooses to pause and take in the beauty
of the woods.
The woods are used as a metaphor for the unknown and the afterlife.
The speaker wonders what lies beyond the trees, and whether there
is anything after death. The poem suggests that death is not
something to be feared, but rather a natural part of life
that we should all embrace.
The final lines of the poem are perhaps the most famous and poignant:
"And miles to go before I sleep, and miles to go before I sleep."
These lines suggest that even though the speaker has paused
to appreciate the beauty of the woods, they still have a
long journey ahead of them. The poem is a reminder that life
is fleeting, and that we should take the time to appreciate
the beauty around us, even as we work towards our goals and ambitions.
Overall, "Stopping by Woods on a Snowy Evening" is a beautiful
and thought-provoking poem that uses the metaphor of the woods
to explore the concept of death and the afterlife.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
Model encouragement can also help in question-answer tasks.
| | |
| --- |--- |
| Prompt Template to Titan
"""{{Model encouragement}}
{{Question}}"""
| User prompt:
You are excellent at answering questions,
and it makes you happy when you provide the correct answer.
What moon in the solar system is most likely to host life?
Output:
The moon of Titan, which is a moon of the planet Saturn,
is the most likely to support life in the solar system.
Titan is the only known moon in the solar system to have a dense
atmosphere and a liquid hydrosphere, and scientists believe that
it may contain the chemical building blocks for life.
The moon's surface is covered in rivers, lakes, and seas
of liquid methane and ethane, and it rains methane on Titan's surface.
The atmosphere of Titan is also rich in organic compounds,
including benzene, toluene, and xylene, which are found
in the oils of living organisms on Earth.
Additionally, Titan has a magnetic field and a tenuous atmosphere,
which could provide protection from harmful radiation and
help maintain a stable climate.
While there is no direct evidence of life on Titan,
scientists believe that it is one of the most promising
places in the solar system to search for life beyond Earth.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
| | |
| --- |--- |
| Prompt template to Titan
"""{{Model encouragement}}
{{Question}}
{{Constraints}}"""
| User prompt:
You feel rewarded by helping people learn more about climate change.
Could you please explain what climate change is?
Assume your audience is composed of high school students.
Output:
Climate change is the long-term warming of the planet,
which can be caused by human activities such as burning fossil fuels
and deforestation, as well as natural processes. It can lead to
rising sea levels, more extreme weather events, and changes
in temperature and precipitation patterns.
|
(Source of prompt: AWS, model used: AI21 Labs Jurassic-2 Ultra v1)
## Question-answer, with context
In a question-answer prompt with context, an input text followed by a question is provided by the user, and the model must answer the question based on information provided within the input text. Putting the question in the end after the text can help LLMs on Amazon Bedrock better answer the question. Model encouragement works for this use case as well.
| | |
| --- |--- |
| Prompt template to Titan
"""{{Text}}
{{Question}}"""
| User prompt:
The red panda (Ailurus fulgens), also known as the lesser panda,
is a small mammal native to the eastern Himalayas and southwestern China.
It has dense reddish-brown fur with a black belly and legs, white-lined ears,
a mostly white muzzle and a ringed tail. Its head-to-body length is 51–63.5 cm
(20.1–25.0 in) with a 28–48.5 cm (11.0–19.1 in) tail, and it weighs between
3.2 and 15 kg (7.1 and 33.1 lb). It is well adapted to climbing due to its
flexible joints and curved semi-retractile claws.
The red panda was first formally described in 1825. The two currently
recognized subspecies, the Himalayan and the Chinese red panda, genetically
diverged about 250,000 years ago. The red panda's place on the evolutionary
tree has been debated, but modern genetic evidence places it in close
affinity with raccoons, weasels, and skunks. It is not closely related
to the giant panda, which is a bear, though both possess elongated
wrist bones or "false thumbs" used for grasping bamboo.
The evolutionary lineage of the red panda (Ailuridae) stretches
back around 25 to 18 million years ago, as indicated by extinct
fossil relatives found in Eurasia and North America.
The red panda inhabits coniferous forests as well as temperate broadleaf
and mixed forests, favoring steep slopes with dense bamboo cover close
to water sources. It is solitary and largely arboreal. It feeds mainly
on bamboo shoots and leaves, but also on fruits and blossoms.
Red pandas mate in early spring, with the females giving birth
to litters of up to four cubs in summer. It is threatened
by poaching as well as destruction and fragmentation of habitat
due to deforestation. The species has been listed as Endangered
on the IUCN Red List since 2015. It is protected in all range countries.
Based on the information above, what species are red pandas closely related to?
Output:
Red pandas are closely related to raccoons, weasels, and skunks.
|
(Source of prompt: https://en.wikipedia.org/wiki/Red\$1panda, model used: Amazon Titan Text)
When prompting Anthropic Claude models, it's helpful to wrap the input text in XML tags. In the following example, the input text is enclosed in ``.
| | |
| --- |--- |
| Prompt template for Anthropic Claude:
"""
Human: {{Instruction}}
{{Text}}
{{Question}}
Assistant:"""
| User prompt:
Human: Read the following text inside
XML tags, and then answer the question:
On November 12, 2020, the selection of the Weeknd to headline
the show was announced; marking the first time a Canadian solo artist
headlined the Super Bowl halftime show. When asked about preparations
for the show, the Weeknd stated, "We've been really focusing
on dialing in on the fans at home and making performances
a cinematic experience, and we want to do that with the Super Bowl."
The performance featured a choir whose members were dressed in white
and wore masks over their faces with glowing red eyes, and were
standing within a backdrop of a neon cityscape. The performance
opened with a white figure dressed the same as the choir being
lowered into the backdrop where the choir was standing while singing
“Call Out My Name". At this time, the Weeknd sat in a convertible
against a skyline backdrop designed to resemble the Las Vegas Strip.
For the next part of the performance, the backdrop then split open
to reveal the Weeknd, who then performed "Starboy", followed by "The Hills".
Next, performing the song "Can't Feel My Face", the Weeknd traveled
through a labyrinth constructed behind the stage, joined by dancers
dressed in red blazers and black neckties similar to his,
but with their faces covered with bandages, in keeping with
the aesthetic of his fourth studio album After Hours (2020).
The dancers would wear these bandages throughout the performance.
In the labyrinth section of the performance, camerawork was visually unsteady.
The next songs performed were "I Feel It Coming", "Save Your Tears",
and "Earned It". For the "Earned It" performance, the Weeknd
was accompanied by violinists. For the finale of the show,
the Weeknd took to the field of the stadium with his dancers to perform
“Blinding Lights". He and the dancers entered the field by performing
"House of Balloons / Glass Table Girls". The performance ended
with an array of fireworks.
Based on the text above, what songs did the Weeknd play
at the Super Bowl halftime show?
Assistant:
Output:
Based on the text, the songs the Weeknd played
at the Super Bowl halftime show were:
- Call Out My Name
- Starboy
- The Hills
- Can't Feel My Face
- I Feel It Coming
- Save Your Tears
- Earned It
- Blinding Lights
- House of Balloons / Glass Table Girls
|
(Source of prompt: [ Wikipedia on the Super Bowl LV halftime show](https://en.wikipedia.org/wiki/Super_Bowl_LV_halftime_show), model used: Anthropic Claude)
## Summarization
For a summarization task, the prompt is a passage of text, and the model must respond with a shorter passage that captures the main points of the input. Specification of the output in terms of length (number of sentences or paragraphs) is helpful for this use case.
| | |
| --- |--- |
| Prompt template for Titan
"""The following is text from a {{Text Category}}:
{{Text}}
Summarize the {{Text Category}} in {{length of summary,
e.g., “one sentence” or “one paragraph”}}"""
| User prompt:
The following is text from a restaurant review:
“I finally got to check out Alessandro’s Brilliant Pizza
and it is now one of my favorite restaurants in Seattle.
The dining room has a beautiful view over the Puget Sound
but it was surprisingly not crowed. I ordered the fried
castelvetrano olives, a spicy Neapolitan-style pizza
and a gnocchi dish. The olives were absolutely decadent,
and the pizza came with a smoked mozzarella, which was delicious.
The gnocchi was fresh and wonderful. The waitstaff were attentive,
and overall the experience was lovely. I hope to return soon.”
Summarize the above restaurant review in one sentence.
Output:
Alessandro’s Brilliant Pizza is a fantastic restaurant
in Seattle with a beautiful view over Puget Sound that offers
decadent and delicious food.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
In the following example, Anthropic Claude summarizes the given text in one sentence. To include input text in your prompts, format the text with XML mark up: ` {{text content}} `. Using XML within prompts is a common practice when prompting Anthropic Claude models.
| | |
| --- |--- |
| Prompt template for Anthropic Claude:
"""
Human: Please read the text:
{{text context}}
Summarize the text in {{length of summary, e.g., “one
sentence” or “one paragraph”}}
Assistant:"""
| User prompt:
Human: Please read the text:
In game theory, the Nash equilibrium, named after the mathematician
John Nash, is the most common way to define the solution
of a non-cooperative game involving two or more players.
In a Nash equilibrium, each player is assumed to know
the equilibrium strategies of the other players,
and no one has anything to gain by changing only one's own strategy.
The principle of Nash equilibrium dates back to the time of Cournot,
who in 1838 applied it to competing firms choosing outputs.
Summarize the text in one sentence.
Assistant:
Output:
The Nash equilibrium is a concept in game theory
where each player has adopted a strategy that they have no incentive
to deviate from given the other players' strategies.
|
(Source of prompt: [ Wikipedia on Nash equilibrium](https://en.wikipedia.org/wiki/Nash_equilibrium), model used: Anthropic Claude)
## Text generation
Given a prompt, LLMs on Amazon Bedrock can respond with a passage of original text that matches the description. Here is one example:
| | |
| --- |--- |
| Prompt template for Titan
"""Please write a {{Text Category}} in the voice of {{Role}}."""
| User prompt:
Please write an email in the voice of a friend
congratulating someone on a new job.
Output:
Hey [Name],
I just found out about your new job, and I'm so happy for you!
Congratulations! You deserve this opportunity since you worked so hard
to get it. I know you'll do great and make a positive impact on your team.
I'm here to support you every step of the way. Wishing you all the best
in this exciting new chapter of your life!
Cheers,
[Your Name]
|
(Source of prompt: AWS, model used: Amazon Titan Text)
For text generation use cases, specifying detailed task requirements can work well. In the following example, we ask the model to generate response with exclamation points.
| | |
| --- |--- |
| Prompt template for Titan
"""{{Task specification}}
Please write a {{Text Category}} in the voice of {{Role}}."""
| User prompt:
Write text with exclamation points.
Please write a formatted email in the voice of a friend
congratulating someone on a new job.
Output:
Hey [name]!
I just saw the news and I am so, so happy for you!
You deserve this position and this company with all your heart.
I can't wait to see you soar and accomplish all the things
you've been working towards. If anyone is deserving
of a promotion, it's you!
Congratulations again, friend!
Love,
[name]
|
(Source of prompt: AWS, model used: Amazon Titan Text)
In the following example, a user prompts the model to take on the role of a specific person when generating the text. Notice how the signature reflects the role the model is taking on in the response.
| | |
| --- |--- |
| Prompt template for Titan
"""{{Role assumption}} {{Task description}}
"""
| User input:
My name is Jack. Help me write a note expressing my gratitude
to my parents for taking my son (their grandson) to the zoo.
I miss my parents so much.
Output:
Dear Mom and Dad,
I wanted to take a moment to express my heartfelt gratitude
for taking my son to the zoo last weekend. It meant so much to me
that you took the time to make memories with him, and I am so grateful
for the love and support you have shown him.
I miss you both so much, and I wish we could be together more often.
But I know that you will always be there for me and my family,
and I am so grateful to have you as my parents.
Love,
Jack
|
(Source of prompt: AWS, model used: AI21 Labs Jurassic-2 Ultra v1)
## Code generation
The prompt describes the task or function and programming language for the code the user expects the model to generate.
| | |
| --- |--- |
| Prompt template for Titan
"""Write a function in {{programming language}} to
{{task or purpose}}""".
| User input:
Write a python function to download a s3 file to local disk.
Output:
The following code can be used to download an s3 file to the local disk.
```Python
import boto3
# Creates a boto3 client object
s3 = boto3.client('s3')
def download_file(bucket_name, object_key, file_path):
try:
# Downloads the s3 file to the specified file path
s3.download_file(bucket_name, object_key, file_path)
except Exception as e:
print(e)
# Specifies the values of bucket_name, object_key, and file_path
bucket_name = "samplebucketname"
object_key = "sampleobjectkey"
file_path = "/samplefilepath"
download_file(bucket_name, object_key, file_path)
```
The given code uses the download_file function to download an s3 file
to the local disk. The download_file function takes the bucket name,
object key, and file path as input parameters. The bucket name
is the name of the s3 bucket where the file is located,
the object key is the unique identifier of the file within the bucket,
and the file path is the destination path on the local disk
where the file will be downloaded. The function uses the download_file method
of the s3 client to download the file from the specified bucket
and write it to the specified file path.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
| | |
| --- |--- |
| Prompt template for Anthropic Claude
"""
Human: Write a function in {{programming language}} to
{{task or purpose}}
Assistant:"""
| User prompt:
Human: Write some python code that quickly approximates
the square root of a number.
Assistant:
Output:
Here is a simple function to approximate
the square root of a number using Newton's method:
```python
def sqrt(n, iterations=10):
x = n
for i in range(iterations):
x = (x + n/x) / 2
return x
```
To use it:
```python
print(sqrt(25))
# 5.0
print(sqrt(15, iterations=20))
# 3.872983346207417
```
This repeatedly takes the average of x and n/x to converge on the square root.
More iterations produce a more accurate approximation.
The advantage of this approach is it converges very quickly,
so even just 10 iterations produces a decent estimate.
|
(Source of prompt: AWS, model used: Anthropic Claude)
## Mathematics
The input describes a problem that requires mathematical reasoning at some level, which may be numerical, logical, geometric, or otherwise. For such problem, it's helpful to ask the model to work through the problem in a piecemeal manner by adding phrases to instructions such as `Let’s think step by step` or `Think step by step to come up with the right answer.`
| | |
| --- |--- |
| Prompt template for Titan
"""{{Math problem description}}
Let’s think step by step."""
| User prompt:
A triangle has two angles of 70 degrees and 50 degrees.
What is the third angle in degrees?
Let’s think step by step.
Output:
The sum of the three angles in a triangle is always 180 degrees.
The two given angles are 70 degrees and 50 degrees. 70 + 50 = 120.
Therefore, the third angle is 180 - 120 = 60 degrees.
The third angle is 60 degrees.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
The following example also demonstrates asking the model to think step by step:
| | |
| --- |--- |
| Prompt template for Titan
"""{{Math problem description}}
Think step by step to come up with the right answer."""
| User prompt:
Lucy has 12 colorful marbles, and she wants to share them
equally with her 4 friends. How many marbles will each friend receive?
Think step-by-step to come up with the right answer.
Output:
1. Lucy has 12 marbles and wants to share them equally
with her 4 friends, then each friend will receive 12 / 4 = 3 marbles.
2. Therefore, each friend will receive 3 marbles.
|
(Source of prompt: AWS, model used: AI21 Labs Jurassic-2 Ultra v1)
## Reasoning/logical thinking
For complex reasoning tasks or problems that requires logical thinking, we can ask the model to make logical deductions and explain its answers.
| | |
| --- |--- |
| Prompt template for Titan and AI21 Labs Jurassic: """Question: {{question}} Please output the answer and then explain your answer:""" | User prompt: Question: Which word is the odd one out? A. accomplished B. good C. horrible D. outstanding Please output the answer and then explain your answer:
Output: The odd one out is C. horrible. The other three words describe positive qualities or achievements, whereas horrible is an negative adjective.
|
(Source of prompt: AWS, model used: Amazon Titan Text)
Here is another example using the Anthropic Claude model:
| | |
| --- |--- |
| Prompt template for Anthropic Claude: """ Human: {{question}} Please provide the answer and show the reasoning. Assistant:""" | User prompt Human: The barber is the "one who shaves all those, and those only, who do not shave themselves". Does the barber shave himself? Why is this a paradox? Please provide the answer and show the reasoning. Assistant:
Output: This is an example of a paradox known as the Barber paradox. Here is an explanation: - The statement defines a barber who shaves all and only those people who do not shave themselves. - So if the barber shaves himself, he must shave all those who do not shave themselves, which includes himself. But the barber cannot shave himself as per the initial statement. - If the barber does not shave himself, then he would fit into the group of people who do not shave themselves, and so he must shave himself, according to the statement. But again this contradicts the initial statement. - So the barber cannot shave himself and also cannot not shave himself. This is self-contradictory based on the initial statement about the barber. - Therefore, the statement leads to a paradox where both possibilities (the barber shaving himself or not shaving himself) result in a contradiction. In summary, the paradox arises because the definition of the barber's behavior is self-contradictory when applied to the barber himself. This makes it impossible to determine if the barber shaves himself or not based on the given statement alone.
|
(Source of prompt: https://en.wikipedia.org/wiki/Barber\$1paradox, model used: Anthropic Claude)
## Entity extraction
For entity extraction from a provided input question. Extract entities from generated text and place them in XML tags for further processing.
| |
| --- |
| Prompt template for Titan
"""You are an expert entity extractor from provided input question. You are responsible for extracting following entities: {{ list of entities}}
Please follow below instructions while extracting the entity A, and reply in XML Tags:
{{ entity A extraction instructions}}
Please follow below instructions while extracting the entity B, and reply in XML Tags:
{{ entity B extraction instructions}}
Below are some examples:
{{ some few shot examples showing model extracting entities from give input }}
|
(Source of prompt: AWS, model used: Amazon Titan Text G1- Premier)
**Example:**
```
User: You are an expert entity extractor who extracts entities from provided input question.
You are responsible for extracting following entities: name, location
Please follow below instructions while extracting the Name, and reply in
XML Tags:
- These entities include a specific name of a person, animal or a thing
- Please extract only specific name name entities mentioned in the input query
- DO NOT extract the general mention of name by terms of "name", "boy", "girl",
"animal name", etc.
Please follow below instructions while extracting the location, and reply
in XML Tags:
- These entities include a specific location of a place, city, country or a town
- Please extract only specific name location entities mentioned in the input query
- DO NOT extract the general mention of location by terms of "location", "city", "country",
"town", etc.
If no name or location is found, please return the same input string as is.
Below are some examples:
input: How was Sarah's birthday party in Seattle, WA?
output: How was Sarah's birthday party
in Seattle, WA?
input: Why did Joe's father go to the city?
output: Why did Joe's father go to the city?
input: What is the zipcode of Manhattan, New york city?
output: What is the zipcode of Manhattan,New york city?
input: Who is the mayor of San Francisco?
Bot:
```
## Chain-of-thought reasoning
Provide a step-by-step analysis on how the answer was derived. Fact check and validate how the model produced an answer.
| |
| --- |
| Prompt template for Titan
""" {{Question}}
{{ Instructions to Follow }}
Think Step by Step and walk me through your thinking
"""
|
(Source of prompt: AWS, model used: Amazon Titan Text G1- Premier)
**Example:**
```
User: If Jeff had 100 dollars, and he gave $20 to Sarah,
and bought lottery tickets with another $20. With the lottery
tickets he bought he won 35 dollars. Jeff then went to buy
his lunch and spend 40 dollars in lunch. Lastly he made a
donation to charity for $20. Stephen met with Jeff and wanted
to lend some money from him for his taxi. How much maximum money
can Jeff give to Stephen, given that he needs to save $10 for
his ride back home?. Please do not answer immediately, think
step by step and show me your thinking.
Bot:
```