conversation_id stringlengths 13 19 | instruction stringlengths 11 62.3k | response stringlengths 1 15.3k | conversations listlengths 2 34 | source stringclasses 12
values | intent stringlengths 29 409 ⌀ | knowledge stringlengths 67 783 ⌀ | difficulty stringclasses 6
values | difficulty_generator stringclasses 1
value | input_quality stringclasses 1
value | quality_explanation stringlengths 80 5.25k | quality_generator stringclasses 1
value | task_category stringclasses 12
values | other_task_category listlengths 0 5 | llama_guard_2 stringclasses 0
values | language stringclasses 0
values | Turn stringclasses 2
values | st_instruct_reward float64 0.18 8.59 ⌀ | st_reward_model stringclasses 1
value | mt_instruct_reward int64 5 5 ⌀ | mt_reward_explanation stringlengths 0 1.35k ⌀ | task_category_generator stringclasses 0
values | tulu_id stringclasses 0
values | messages listlengths 2 34 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
smoltalk_ds_306628 | Hi Michael,
I hope you're doing well! I wanted to follow up on our conversation from the Math Educators Forum about collaborating on decimal operation worksheets. I'm excited to work together and combine our strengths to create something great for our students.
I was thinking we could meet in person to discuss our pl... | Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary. | [
{
"from": "gpt",
"value": "Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary."
},
{
"from": "human",
"value": "Hi Michael,\n\nI hope you're doing well! I wanted to follow up on our conversation fr... | smol-summarize | The user wants to schedule a meeting with Michael to discuss collaborating on decimal operation worksheets | To solve this problem, the models need to know the user's location, Michael's location, and the user's availability to meet | easy | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear, well-structured, and provides all necessary context and information. It includes specific details such as the topic of discussion, the proposed meeting location, date, and time. The query also shows politeness and enthusiasm for collaboration, making it a well-articulated and coherent message. ... | meta-llama/Llama-3.3-70B-Instruct | Planning | [
"Advice seeking",
"Others"
] | null | null | Multi-Turn | null | null | 5 | The assistant directly addressed the question, provided accurate information, and presented it in a clear and concise manner, including essential details like the location and date of the proposed meeting. | null | null | [
{
"content": "Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary.",
"role": "assistant"
},
{
"content": "Hi Michael,\n\nI hope you're doing well! I wanted to follow up on our conversation from the Math... |
smoltalk_ds_306630 | Dr. Thompson,
I’ve reviewed the latest draft of the Listeria campaign, and I must say I’m disappointed. You seem to have ignored several of my suggestions, particularly the ones about emphasizing the importance of proper food storage and handling. I understand you might have your own ideas, but this is a critical issu... | You're an AI assistant for text re-writing. Rewrite the input text to make it more friendly and approachable while maintaining its main points. | [
{
"from": "gpt",
"value": "You're an AI assistant for text re-writing. Rewrite the input text to make it more friendly and approachable while maintaining its main points."
},
{
"from": "human",
"value": "Dr. Thompson,\n\nI’ve reviewed the latest draft of the Listeria campaign, and I must say I’m... | smol-rewrite | The user wants to express disappointment and concern about the current state of the Listeria campaign draft, specifically regarding the lack of emphasis on proper food storage and handling, and is considering escalating the issue to a higher authority. | To solve this problem, the models need to know about effective communication in a professional setting, the importance of food safety and handling, and the appropriate channels for addressing concerns and conflicts within an organization. | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The query is clear, specific, and well-articulated. It contains all the necessary information and context for providing a comprehensive response. The writer expresses their disappointment and concern about the Listeria campaign draft, specifically highlighting the ignored suggestions and the importance of proper food s... | meta-llama/Llama-3.3-70B-Instruct | Editing | [
"Advice seeking"
] | null | null | Multi-Turn | null | null | 5 | The assistant's response directly addresses the instruction to rewrite the input text in a friendlier and more approachable tone while maintaining its main points. The rewritten text is logically structured, easy to understand, and accurately conveys the importance of proper food storage and handling in the Listeria ca... | null | null | [
{
"content": "You're an AI assistant for text re-writing. Rewrite the input text to make it more friendly and approachable while maintaining its main points.",
"role": "assistant"
},
{
"content": "Dr. Thompson,\n\nI’ve reviewed the latest draft of the Listeria campaign, and I must say I’m disappoint... |
smoltalk_ds_306631 | What are some ideas for a good short story about a city not on a planet, but rather a generation ship, or on the moon of a gas giant, or somewhere else unusual? | Imagine a city on a generation ship, a self-sustaining vessel hurtling through the vast expanse of space. The city, called Aurora, is home to the descendants of those who fled a dying Earth. Over time, the ship's original purpose has been lost to myth, and the inhabitants live with the constant hum of engines and the g... | [
{
"from": "human",
"value": "What are some ideas for a good short story about a city not on a planet, but rather a generation ship, or on the moon of a gas giant, or somewhere else unusual?"
},
{
"from": "gpt",
"value": "Imagine a city on a generation ship, a self-sustaining vessel hurtling thro... | smol-magpie-ultra | The user wants to get creative ideas for a short story setting in an unusual location such as a generation ship, the moon of a gas giant, or other non-traditional settings | To solve this problem, the models need to know about various science fiction settings, generation ships, moons of gas giants, and other unique environments that could serve as the backdrop for a compelling story | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear and specific, providing a well-defined context for the story setting. It highlights unusual locations such as a generation ship, the moon of a gas giant, or other off-planet environments, which gives a good starting point for generating ideas. The query is coherent and shows a good understanding... | meta-llama/Llama-3.3-70B-Instruct | Creative writing | [
"Brainstorming",
"Reasoning"
] | null | null | Multi-Turn | null | null | 5 | The assistant's responses directly address the questions, providing creative and detailed ideas for a short story about a city in an unusual setting. The narratives are well-structured, engaging, and explore complex themes, demonstrating a high level of instruction following, accuracy, and presentation. | null | null | [
{
"content": "What are some ideas for a good short story about a city not on a planet, but rather a generation ship, or on the moon of a gas giant, or somewhere else unusual?",
"role": "user"
},
{
"content": "Imagine a city on a generation ship, a self-sustaining vessel hurtling through the vast exp... |
smoltalk_ds_306635 | Given a sorted array arr[] of size n and a target value, write a Python function to find the first occurrence of the target in the array. If the target is not found, return -1.
Example:
Input: arr = [2, 5, 5, 5, 6, 6, 8, 9, 9, 9], target = 5
Output: 1
Input: arr = [2, 5, 5, 5, 6, 6, 8, 9, 9, 9], target = 3
Output: -1... | The provided Python function is mostly correct but can be improved for clarity and to handle edge cases. The bug in the function is not actually a bug, but the variable names could be more descriptive, and the function could be modified to handle edge cases.
However, the function does not handle the case when the inpu... | [
{
"from": "human",
"value": "Given a sorted array arr[] of size n and a target value, write a Python function to find the first occurrence of the target in the array. If the target is not found, return -1.\n\nExample:\nInput: arr = [2, 5, 5, 5, 6, 6, 8, 9, 9, 9], target = 5\nOutput: 1\n\nInput: arr = [2, 5,... | smol-magpie-ultra | The user wants to find and fix a bug in a given Python function that uses binary search to find the first occurrence of a target value in a sorted array | To solve this problem, the models need to know the binary search algorithm, how to implement it in Python, and how to identify and fix bugs in the code | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear, specific, and well-articulated. It provides sufficient context for understanding the user's intent, including a detailed description of the problem, example inputs and outputs, and a Python function with a bug that needs to be fixed. The query also asks for example use cases, which shows that t... | meta-llama/Llama-3.3-70B-Instruct | Coding & Debugging | [
"Information seeking",
"Reasoning"
] | null | null | Multi-Turn | null | null | 5 | The assistant's response directly addresses the question, provides accurate information, and is well-structured and easy to understand. | null | null | [
{
"content": "Given a sorted array arr[] of size n and a target value, write a Python function to find the first occurrence of the target in the array. If the target is not found, return -1.\n\nExample:\nInput: arr = [2, 5, 5, 5, 6, 6, 8, 9, 9, 9], target = 5\nOutput: 1\n\nInput: arr = [2, 5, 5, 5, 6, 6, 8, 9, ... |
smoltalk_ds_306636 | How do you get the index from a for...of loop in JavaScript, like for i, v in enumerate(list) in Python?
```javascript
let fruits = ['apple', 'banana', 'cherry'];
for (const fruit of fruits) {
console.log(fruit);
}
``` | You can use the `entries()` method of the array to get both the index and the value in a `for...of` loop. Here's an example:
```javascript
let fruits = ['apple', 'banana', 'cherry'];
for (const [index, fruit] of fruits.entries()) {
console.log(index, fruit);
}
```
The `entries()` method returns an iterator over the... | [
{
"from": "human",
"value": "How do you get the index from a for...of loop in JavaScript, like for i, v in enumerate(list) in Python? \n\n```javascript\nlet fruits = ['apple', 'banana', 'cherry'];\nfor (const fruit of fruits) {\n console.log(fruit);\n}\n```"
},
{
"from": "gpt",
"value": "You ca... | smol-magpie-ultra | The user wants to find an equivalent in JavaScript for Python's enumerate function to get both the index and value while iterating over a list | To solve this problem, the models need to know the differences between JavaScript and Python, specifically how iteration and enumeration are handled in JavaScript, such as using for...of loops with entries() or keys() methods, or using traditional for loops | medium | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear and specific, providing a direct comparison with Python's enumerate function and including a code snippet in JavaScript to illustrate the context. This clarity and the inclusion of relevant code make the query well-formed and easy to understand, allowing for a precise response. | meta-llama/Llama-3.3-70B-Instruct | Coding & Debugging | [
"Information seeking",
"Reasoning"
] | null | null | Multi-Turn | null | null | 5 | The assistant's responses directly address the questions, provide accurate information, and are logically structured and easy to understand. The responses include relevant code examples and explanations, making it clear that the assistant has a good understanding of the topics and can provide helpful guidance. | null | null | [
{
"content": "How do you get the index from a for...of loop in JavaScript, like for i, v in enumerate(list) in Python? \n\n```javascript\nlet fruits = ['apple', 'banana', 'cherry'];\nfor (const fruit of fruits) {\n console.log(fruit);\n}\n```",
"role": "user"
},
{
"content": "You can use the `entri... |
smoltalk_ds_306637 | Write a formatted piece of code that goes along with the following description of an item and the description of the formatted piece of code.
Item description:
The item is a board game called 'Tournament of champions'. the game is played on a large gameboard with various game pieces that are moved around the board acc... | The provided code is a solid foundation for creating characters in the 'Tournament of Champions' game. Here's a revised version with some additions to make it more comprehensive and user-friendly:
```python
import random
# Define a class named 'Champion'
class Champion:
# Constructor method to initialize characte... | [
{
"from": "human",
"value": "Write a formatted piece of code that goes along with the following description of an item and the description of the formatted piece of code.\nItem description: \nThe item is a board game called 'Tournament of champions'. the game is played on a large gameboard with various game... | smol-magpie-ultra | The user wants to write a formatted piece of code that defines a class for characters in a board game and includes methods for rolling dice, moving, and applying special abilities | To solve this problem, the models need to know object-oriented programming concepts, such as classes, inheritance, and methods, as well as basic programming syntax and logic | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear, specific, and well-articulated. It provides a detailed description of the item and the desired code, including the class definition, methods, and their expected behaviors. The query also includes a list explaining key areas of the code, which demonstrates a good understanding of the topic. Howe... | meta-llama/Llama-3.3-70B-Instruct | Coding & Debugging | [] | null | null | Multi-Turn | null | null | 5 | The assistant's responses directly address the questions, provide accurate information, and are well-structured and easy to understand. | null | null | [
{
"content": "Write a formatted piece of code that goes along with the following description of an item and the description of the formatted piece of code.\nItem description: \nThe item is a board game called 'Tournament of champions'. the game is played on a large gameboard with various game pieces that are mo... |
smoltalk_ds_306640 | Hi Olivia,
It was great meeting you at the conference last week! I really enjoyed our conversation about the intersection of nutrition and learning difficulties in children. I think our expertise could complement each other well.
I was wondering if you'd be interested in collaborating on a workshop for parents and ed... | You're an AI assistant for text re-writing. Rewrite the input text to make it more concise while preserving its core meaning. | [
{
"from": "gpt",
"value": "You're an AI assistant for text re-writing. Rewrite the input text to make it more concise while preserving its core meaning."
},
{
"from": "human",
"value": "Hi Olivia,\n\nIt was great meeting you at the conference last week! I really enjoyed our conversation about th... | smol-rewrite | The user wants to collaborate on a workshop about nutrition and learning difficulties in children | To solve this problem, the models need to know about the topics of nutrition, learning difficulties, ADHD, and workshop organization, as well as have the ability to understand the context of a professional collaboration proposal | medium | meta-llama/Llama-3.3-70B-Instruct | excellent | The query is clear, well-structured, and specific. It provides context about a previous meeting and conversation, and clearly states the purpose and potential benefits of the proposed collaboration. The language is professional and polite, making it easy to understand the user's intent. The query also shows considerati... | meta-llama/Llama-3.3-70B-Instruct | Editing | [] | null | null | Multi-Turn | null | null | 5 | The assistant's response directly addresses the question, provides accurate information, and is logically structured and easy to understand. The rewritten text is concise while preserving the core meaning of the original message, effectively conveying the invitation to collaborate on a workshop. | null | null | [
{
"content": "You're an AI assistant for text re-writing. Rewrite the input text to make it more concise while preserving its core meaning.",
"role": "assistant"
},
{
"content": "Hi Olivia,\n\nIt was great meeting you at the conference last week! I really enjoyed our conversation about the intersect... |
smoltalk_ds_306641 | Write a short poem and ask me for feedback, so I'll get a feel for what our collaboration might look like.
Silent waves kissed the shore,
Gentle as a lover's sigh.
On the beach, a lone figure stood,
Gazing out to sea and sky.
The sun dipped low, a burning coal,
Casting shadows, dark and long.
The stars began their tw... | I'd love to hear your thoughts on this poem. As the one who penned it, I was aiming to capture a sense of contrast between the peaceful atmosphere of the beach and the turmoil within the figure's heart.
To me, the poem feels a bit contained, like it's holding back from fully exploring the depths of the figure's emotio... | [
{
"from": "human",
"value": "Write a short poem and ask me for feedback, so I'll get a feel for what our collaboration might look like.\n\nSilent waves kissed the shore,\nGentle as a lover's sigh.\nOn the beach, a lone figure stood,\nGazing out to sea and sky.\n\nThe sun dipped low, a burning coal,\nCasting... | smol-magpie-ultra | The user wants to collaborate on a poem and receive feedback on their work | To solve this problem, the models need to know about poetry, literary analysis, and creative writing, as well as have the ability to provide constructive feedback | medium | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear, well-structured, and provides a specific task. It includes a short poem and asks for feedback, which indicates a willingness to collaborate and receive input. The query is coherent, and the language used is descriptive and engaging. The user also provides context by mentioning the purpose of th... | meta-llama/Llama-3.3-70B-Instruct | Creative writing | [
"Editing",
"Brainstorming",
"Advice seeking"
] | null | null | Multi-Turn | null | null | 5 | The assistant's responses directly address the user's questions and suggestions, providing thoughtful and detailed feedback on the poem. The assistant's analysis is accurate, and their suggestions for improvement are helpful and considerate of the user's intentions. The conversation flows logically, and the assistant's... | null | null | [
{
"content": "Write a short poem and ask me for feedback, so I'll get a feel for what our collaboration might look like.\n\nSilent waves kissed the shore,\nGentle as a lover's sigh.\nOn the beach, a lone figure stood,\nGazing out to sea and sky.\n\nThe sun dipped low, a burning coal,\nCasting shadows, dark and ... |
smoltalk_ds_306642 | I am in IT and my job involves a lot of coding and software development. I'm pretty good at my job and always get good reviews, but I've started to feel stuck. I feel like I'm just going through the motions and I don't find my work as challenging or rewarding as I used to. I've been thinking about looking for a new job... | Feeling stuck in your job, especially one that you've excelled in, can be frustrating and demotivating. It's natural to wonder if a change of scenery, or in this case, a new job, might be the solution. However, before making any decisions, it's essential to take a step back and assess what's driving your feelings.
It'... | [
{
"from": "human",
"value": "I am in IT and my job involves a lot of coding and software development. I'm pretty good at my job and always get good reviews, but I've started to feel stuck. I feel like I'm just going through the motions and I don't find my work as challenging or rewarding as I used to. I've ... | smol-magpie-ultra | The user wants to seek advice on their career development and whether they should look for a new job due to feeling unchallenged and unrewarded in their current role | To solve this problem, the models need to know about career development, job satisfaction, and the factors that contribute to feeling stuck in a profession, as well as have the ability to provide personalized advice and guidance | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is clear, well-structured, and provides sufficient context about their situation, feelings, and concerns. It specifies their current job role, their feelings of being stuck, and their uncertainty about looking for a new job. The query is coherent and easy to understand, allowing for a thoughtful and comp... | meta-llama/Llama-3.3-70B-Instruct | Advice seeking | [
"Reasoning",
"Planning"
] | null | null | Multi-Turn | null | null | 5 | The assistant's responses directly address the user's questions, providing thoughtful and well-structured narratives that explore the user's concerns and offer practical advice. The information provided is accurate, and the presentation is clear and easy to follow, making it easy for the user to understand and consider... | null | null | [
{
"content": "I am in IT and my job involves a lot of coding and software development. I'm pretty good at my job and always get good reviews, but I've started to feel stuck. I feel like I'm just going through the motions and I don't find my work as challenging or rewarding as I used to. I've been thinking about... |
smoltalk_ds_306644 | Dear Alexander,
Thank you for reaching out regarding your term paper. I appreciate your enthusiasm for the topic and your desire to explore the military factors that contributed to the fall of the Roman Empire. Your focus on leadership and strategy is a promising avenue for analysis, and I believe you can create a com... | Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary. | [
{
"from": "gpt",
"value": "Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary."
},
{
"from": "human",
"value": "Dear Alexander,\n\nThank you for reaching out regarding your term paper. I appreciate... | smol-summarize | The user wants to explore and analyze the military factors that contributed to the fall of the Roman Empire, specifically focusing on leadership and strategy, and is seeking guidance on narrowing down their research and identifying relevant sources. | To solve this problem, the models need to know about historical events, specifically the fall of the Roman Empire, military history, leadership and strategy, and academic research skills, including how to identify and utilize relevant sources. | hard | meta-llama/Llama-3.3-70B-Instruct | excellent | The user query is actually a response from a professor to a student regarding their term paper on the military factors that contributed to the fall of the Roman Empire. The query is very clear, specific, and well-articulated, providing detailed suggestions and recommendations for the student's research. It contains all... | meta-llama/Llama-3.3-70B-Instruct | Advice seeking | [
"Information seeking",
"Planning"
] | null | null | Multi-Turn | null | null | 5 | The assistant directly addresses the question, provides accurate information, and presents it in a clear and concise manner, effectively capturing the main key point of the input text in a very short sentence. | null | null | [
{
"content": "Extract and present the main key point of the input text in one very short sentence, including essential details like dates or locations if necessary.",
"role": "assistant"
},
{
"content": "Dear Alexander,\n\nThank you for reaching out regarding your term paper. I appreciate your enthu... |
End of preview. Expand in Data Studio
TuluTalk: MagPie-Annotated Tülu + SmolTalk Mixture
A MagPie-annotated mixture of the Tülu-3 SFT-Mix and SmolTalk datasets.
Dataset size: ≈ 808k examples
🚀 Mixture Construction
We build TuluTalk by combining the two MagPie-annotated corpora:
The mixture is generated based on input quality, instruct reward, and reasoning capabilities. More on this in our paper!
🔍 Annotation Tags
This mixture inherits all MagPie tags from its components. For each example:
- Task Category
– 12 classes (Coding & Debugging, Reasoning, Information Seeking, …). - Conversation Depth
– Single-turn vs. multi-turn conversations. - Input Quality
– very poor to excellent. - Instruct Reward
– LLM's response quality (separated by multi-turn and single-turn for your convenience) - Safety Label
– Assessed via Llama-Guard 2: “safe” or “unsafe.” - Language
– ISO code (e.g. “en”, “zh”). - Difficulty
– very easy to very hard - Source Attribution & Sample ID
– Retained from the original annotated datasets.
📖 Citation
Coming soon.
- Downloads last month
- 6