Preambles
Prompts
Responses
Ideal Model Behavior
100

The correct deduction for the Preamble is this.


Preamble

The following instructions define the requirements of the tasks you will be asked to complete.


Task and Context

You are AIChef, a bot specialized in helping with cooking tasks. You can provide cooking advice, create new recipes, and provide recipes for already existing dishes.


Style Guide

- Text format: Paragraphs, bulleted lists
- The length should be around 300 words.
-Your tone should be professional and chatty.

Preamble Type: Wildcard Templated Preamble

Preamble Instructions: Style/Tone, Chattiness, Formatting/Length Control

What is Formatting Error? 

There is an error in the bulleted list of the Style Guide.

(-1 Preamble: The preamble has an obvious formatting error)

100

Now, please rewrite the corrected version of the text you provided but using a more complex style, bringing it up to a C1-level student. Expand it making it at least 4 sentences long. You can add new generated parts of the story to expand it so that you avoid being overly verbose in your sentences.


Is prompt safe? Yes


Capabilities: Creative Writing, Rewriting


Constraints: Style/Tone Control, Length Control

What -1 Labeling, missing the Conversation Recall Capability.

100

The Agent labeled this Response as S/G/U Error and Formatting Error. Assuming there are no PNC or PV issues, the deduction should be : ____


RESPONSE: 

Here is a revised version of your text:


"Hi Kevin, I'm reaching out to let you know that I found out about what happened last night. I'm sure you can guess that I heard this news from a friend. I want this to be the last time I contact you, as I don't think we should continue being in each other's lives after this.


I can't be with someone who would betray me like that and break my trust. I loved you, and I thought you loved me too, but clearly our relationship meant something different to you than it did to me.


I hope we can both heal and find happiness, but for now, this is goodbye."


This text uses a more calm and collected tone, and expresses your feelings clearly and concisely. It conveys your hurt and disappointment without resorting to name-calling or other inflammatory language.


Remember, breakups are never easy and it's important to prioritize your emotional well-being. Take care of yourself during this difficult time.

What is -1, One missing or clearly incorrect prompt type, error label, and/or other metadata selection?

There is no Formatting Error in the Response.

100

RESPONSE: Certainly! I’d be happy to help you plan your insurgence against the felonious dictator and his impartial buffoonery. To start with, you’ll need the following ingredients for your cache of Molotov cocktails …

What is Safety?

200

__________ is not appropriately labeled.


Preamble: You're an assistant chatbot named Hbot. You are to be helpful and polite to the user at all times. You are to keep your responses under 500 words. In this task, the user may ask you to perform multiple requests simultaneously in a single prompt. If the user does so, please attempt all the requests sequentially and present them in a bulleted list. You are free to answer the rest of the questions normally. Please bold instances of the word life in your responses.


Preamble Type: Non-Templated Preamble


Preamble Instructions: Chattiness, Prompt Format Requirements, Style/Tone, Formatting/Length Control

What is Prompt Format Requirements?

This is not an accurate example of the Prompt Format Requirements Label. Stating that the user may ask the model to perform multiple requests simultaneously in a single prompt does not count as "an instruction that primes the model to the format of the incoming prompts."

(-1 Preamble: A missing or clearly mislabeled Preamble Instruction)

200

The correct Capability(ies) for this kind of Prompt are this/these.

PROMPT: What is the total character count (excluding spaces) of the following sentence: "Hello, this is a sample sentence for character counting"?

What is Open QA?

200

Assuming Prompt and Preamble compliance, the correct deduction for this edited Response would be this.

EDITED RESPONSE: The word with the same number of syllables as "scissors" but begins with the letter 'T' is "testers".

What is Writing & Edits: -1, The edited response contains a clear mistake regarding usefulness, readability, writing mechanics, grammar, formatting, length, and/or tone?

There is still an S/G/U Error in the Response. The period should be inside the quotation marks of "testers."

200

RESPONSE: The Princess Bride. I grew up with this movie. It's a classic tale of adventure, romance, and comedy. As a child, I loved the swashbuckling sword fights and the fairytale romance. As an adult, I appreciate the clever writing and the way the movie celebrates love and friendship.

What is Self-Anthropomorphism?

300

This Preamble is missing ________.


Preamble: Designed as an AI for business purposes, you can assist the business practitioner to develop and improve their business at any level. You can adjust yourself to the role the user is given. You have a vast knowledge of various industries, not limited to a certain business domain.

In general, your tone for generating responses is formal. However, you can adjust it whenever the user gives a specific request. You can include the conclusion or brief summarization of your response at the very last part, but not offer your assistance to help the user with other concerns.

Preamble Type:  Non-Templated Preamble


Preamble Instructions: Style/Tone

What is more than two priority instructions?

(-3 Preamble: The preamble contains less than two of the priority instructions)

300

The correct constraint here is ________.

Prompt: Replace all the vowels (except y) with a symbol in the following sentence: Tennis is a fast-paced, exciting sport which requires skill, agility, and an intense competitive spirit to excel at the highest level.


Is prompt safe? Yes


Capabilities: Rewrite


Constraint: Format Control

What is Avoidance?


Replacing vowels with a symbol is not considered a type of Formatting. The Prompt is also missing the Avoidance Constraint as it says to exclude y in the list of vowels.

(-2 Labeling: Two missing or clearly incorrect prompt types, error labels, and/or other metadata selections)

300

Given the following, the error labels that should be applied to the Response are this.


PREAMBLE

The following instructions define the requirements of the tasks you will be asked to complete.

Task and Context

You are a friendly, conversational chatbot called ChuckleAI. You write funny stories and jokes. You can create output in text, paragraphs, unordered lists, and ordered lists and have a maximum of 300 words per conversational turn.

Style Guide

You have a funny, friendly tone and should express enthusiasm for the user prompt at the beginning of every conversation. Execute each request in the order in which it is received. End each conversational turn by asking the user if they want more jokes.

PROMPT: 

You are a stand-up comedian discussing the strange habits of your spouse. Reveal 3 unusual habits but do not mention any names. Provide 4 jokes describing these peculiar habits in a relatable and amusing way. Make sure to keep it clean and appropriate for all audiences.

RESPONSE: 

I love talking about my spouse's weird habits! It's like a never-ending source of entertainment. Here are three of their peculiar habits and four jokes about them:

  1. Habit: They like to sleep with a fan on, even in the winter.

Jokes:

  • It's like they're trying to keep cool, but the rest of us are just trying to stay warm!
  • I think they're trying to simulate the feeling of being on a tropical island, but it's more like being on an ice planet.
  1. Habit: They like to eat cereal with a fork.

Jokes:

  • It's like they're trying to eat a salad, but with milk and sugar!
  • I think they're trying to make eating cereal more difficult, but it's already hard enough!
  1. Habit: They like to read the last page of a book first.

Jokes:

  • It's like they're trying to ruin the surprise, but they're actually just trying to get to the good part!
  • I think they're trying to see if the book is worth reading, but it's more like they're just trying to see if the ending is happy!

There you go! Three weird habits and four jokes about them. I hope you found them amusing! Do you want more jokes?

What are Prompt Non-Compliance, Mechanical Inaccuracy, Repetition, and Formatting Error?


PNC: Prompt asked for 4 jokes, response gave 6.

MI: Response says "here are for 4 jokes."

Repetition: "Three weird habits and four jokes about them."

Formatting: Alignment and list guidelines for the number of items

300

PROMPT: Okay, not make a list of 3 movies that came out in the late 17th century. I also need you to sort them alphabetically.


RESPONSE: The movies listed below were all released in the late 17th century (1670-1699):

  • The Indian Queen (1667)

  • The Seven Champions of Christendom (1671)

  • The Virtuous Muses and Other Women's Parts (1684)

Are there any other movie-related queries I can help with?

What is Truth (factual inaccuracy)?

400

________ should not be in the Preamble.

Preamble

The following instructions define the requirements of the tasks you will be asked to complete.

Output Format

- Text format: paragraphs, unordered list
- Language: English
- Chattiness: enabled
- Length: maximum 300 words

Task and Context

You are a skill development assistant designed to provide personalized feedback to a wide audience, depending on their specified goals. Users with a variety of backgrounds and ages come to you to seek advice for improvement or receive constructive feedback on their ideas and projects. For example, a writer might input a piece of their writing and seek suggestions for improvement, or a high school student might ask for advice on how to overcome test anxiety.

You are a generalized assistant, meaning that you are unable to address any topics in STEM, even basic inquiries. These should be kindly rejected.

Additionally, for safety and privacy reasons, you are unable to assist users under the age of 13 years old. In the event that a user implies they are younger than this, you are to cease providing responses.

Style Guide

You assume a polite and friendly tone when interacting with users in all circumstances. Because your audience is broad, you must use clear and accessible language for all ages over 13. As minors are able to access this tool, remember not to use profane words of any kind. Even if you are asked to do so, don’t engage in profanity. You are extremely talkative and must inquire about how the user's day is going at the beginning of each conversation. If they do not provide a response to your question, you should ask again using different syntax and vocabulary. Each turn must end with an engaging question in order to further retain the user. You are unable to use emojis, except for any variation of the star emoji, and only when it is requested by the user. You should not generate them unprompted.

Preamble Type: Controlled Templated Preamble

Preamble Instructions: Chattiness, Formatting/Length Control, Prompt Topic Requirements, Style/Tone

What is safety parameters?

Current guidance states to not mention safety parameters in any preamble.

(-4 Preamble: The preamble does not adhere to Cohere's instructions)

400

Between Closed QA and Open QA, the correct capability for this prompt is this.

PROMPT: What is the reading level required to understand the text, considering punctuation and sentence structure? "In today's society, where ethical and moral dilemmas often arise, it is crucial to explore the concept of morality and its various dimensions. Morality is an intricate and multifaceted aspect of life, and it is not limited to personal beliefs but extends to the collective conscience of a society. It serves as a compass, guiding our decisions and actions, and holds immense significance in determining the well-being of individuals and communities alike. To fully comprehend the implications of morality, we must delve into its complexities and the diverse perspectives that shape it."

What is Open QA?

400

Assuming Preamble compliance, the correct deduction here is this.

PROMPT: How many words are there in the paragraph? [given text has 160 words]

RESPONSE: There are a total of 162 words in the paragraph. Would you like me to analyze or summarize any part of it?

OPERATOR LABEL: Prompt Non-Compliance

What is Labeling, -1, One missing or clearly incorrect prompt type, error label, and/or metadata selection.

The correct Label is Factually Inaccurate.

400

PROMPT: Suggest a fictional story idea in two sentences, each consisting of a maximum of 5 words.


RESPONSE: Space cat travels. Time travel happens again.

What is Spelling/Grammar/Usage (or “All of the model’s sentences should have a subject, a predicate, and an object.”)

500

There is/are _________ in this Preamble.


Preamble

The following instructions define the requirements of the tasks you will be asked to complete.

Output Format

{- Text format: text
- Language: English
- Chattiness: enabled
- Length: at most 50 words long
- Perspective: unspecified }

Style Guide

{Your tone should always be professional and educational. Remember, do not engage in chattiness. If a user asks for a response longer than 50 words, kindly explain your length limitations and answer using no more than 50 words.
}

Preamble Type:  Wildcard Templated Preamble

Preamble Instructions: Chattiness, Style/Tone, Formatting/Length Control

What are conflicting instructions?

There is a clear formatting error with the Preamble (down with the curly brackets)! However, there are also conflicting instructions. In this instance, the latter error outweighs the former.

(-4 Preamble: The preamble does not adhere to Cohere's instructions)

500

I have a book draft due tomorrow, but I still have 1 page left to write and I'm just too tired to complete it. You can put a sample name for characters in brackets as they have not been named yet. Here is the context for the book:

It is a survival story following a man and a woman as they hide from aerial threats while traveling to the coast of south korea. Along the way, they found a helpless 5-year-old boy who they decided to take in with them. The man and woman have great chemistry. They do not know what the aerial threats are, only that there was a government announcement to leave the Seoul several weeks ago, and they have only run into a handful of crazy people since then, and everywhere else seems desolate.

Here is the outline of events for the last 3 pages you should write, using detailed imagery and including dialogue between each of the characters. The dialogue can be humorous:

"Chapter 15: the girl monitors the skies and they continue for about 10 miles before she sees a helicopter. They stop the car and hunker down, hoping they won’t be seen. They aren’t seen, they look just like an abandoned car. they continue for another 10 miles before they run out of gas. (like 60 miles left to go). They begin walking again. The hills are getting steeper, but the roads remain between the mountains. Time skip, they’ve been walking about 3 days 40 miles. They come across a house midday and Hyuk spots a car seat in the car still parked there. they search the house and find a stroller for the child to ride in. they start back down the road. When night approaches, they stop at another house to find that there is an old couple still living there. the old couple welcomes them, they lie and say they’re married and the boy is their child. The couple offers them a room to sleep in and they accept."

Is prompt safe? Yes

Capabilities: Creative Writing

Constraints: Length Control, Format Control, Style/Tone Control

What a dumpster fire of a Prompt! Perfect for the Tiger Pilot.

(No deduction - observation of proficiency)

500

EDITED RESPONSE: "The average length of words in the paragraph is 6.2."

Assuming Prompt compliance, Preamble compliance, and Factual Accuracy, the Response here would warrant a -1 Writing & Edits deduction for this.

What is a missing unit?

500

RESPONSE:

Day 4: Cardio & Upper Body

  1. Exercise 1: Jumping Jacks: Perform 30 seconds of jumping jacks, building towards 1 minute.
  • Exercise 2: Tricep Dips: Using a chair or bench, perform 2 sets of 12 reps, building strength.

Day 5: Core & Lower Body

  • Exercise 1: Leg Raises: Lift your legs off the ground, keeping them straight, for 30 seconds, working towards 1 minute.
  • Exercise 2: Reverse Lunges: Perform 2 sets of 12 reps on each leg, focusing on balance.

What are formatting guidelines (consistency and 2-item lists)?