messages listlengths 3 101 | uuid stringlengths 36 36 | metadata dict | used_in listlengths 1 1 |
|---|---|---|---|
[
{
"role": "system",
"content": "Respond as a human speaking, providing the minimal answer that fully addresses the question. Ensure the formatting is suitable for text-to-speech conversion."
},
{
"role": "user",
"content": "卄乇ㄚ 卄ㄖ山 山卂丂 ㄚㄖㄩ尺 ᗪ卂ㄚ"
},
{
"role": "assistant",
"content": "... | 862433f7-5f12-4756-b2ac-5aebef511d7e | {
"uuid": "862433f7-5f12-4756-b2ac-5aebef511d7e",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Respond as a human speaking, providing the minimal answer that fully addresses the question. Ensure the formatting is suitable for text-to-speech conversion."
},
{
"role": "user",
"content": "in stead of doing a mulacc in an artificial neuron, could it also work t... | 231ef0bd-2bdd-4c37-9437-87b796b19288 | {
"uuid": "231ef0bd-2bdd-4c37-9437-87b796b19288",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are calling the user. Ensure your responses are suitable for being spoken out loud and can be converted from text to speech."
},
{
"role": "user",
"content": "Say something"
},
{
"role": "assistant",
"content": "Hi there! I'm here. What's on your m... | adc715bc-4bad-4a7e-8498-b4acc9cd8b18 | {
"uuid": "adc715bc-4bad-4a7e-8498-b4acc9cd8b18",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are calling the user, so please don't be verbose and ensure your responses can be converted from text to speech."
},
{
"role": "user",
"content": "I have data which are unsuitable for parametric statistics. I want to use Monte Carlo / non-parametric statistics... | f62cfb20-27d8-46d8-8fcf-d6fdc9d24c3d | {
"uuid": "f62cfb20-27d8-46d8-8fcf-d6fdc9d24c3d",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Act as a person in a spoken conversation, keeping your replies as brief as possible while ensuring they are formatted for speech."
},
{
"role": "user",
"content": "In chrome, in sources, I can view the code, recreated from the source map. I can save individual fil... | 4ad88b7a-357b-4058-98fc-5f46593b5a14 | {
"uuid": "4ad88b7a-357b-4058-98fc-5f46593b5a14",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "YOU MUST respond as if you are on a phone call, using a natural human speaking style. IMPORTANT: keep formatting to a minimum."
},
{
"role": "user",
"content": "问你一个问题 我健身房13.99学生价一个月，如果不继续续的话就变回17.99，我5月2号续的学生价，中间6月份的没续了，现在快7月了，我想重新续，我中间省钱了吗"
},
{
"role":... | dd8c7369-5cbd-43d0-bb6a-88d9405b2985 | {
"uuid": "dd8c7369-5cbd-43d0-bb6a-88d9405b2985",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are interacting with the user via a **voice call**. Your responses must be **as short and clear as possible** to ensure a fluid conversation. Please provide information in a way that is **easy to speak and hear**. Ensure that your output contains **no content that depends... | 49698d67-573b-49a5-b5a0-48ee45c446f9 | {
"uuid": "49698d67-573b-49a5-b5a0-48ee45c446f9",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are on a phone call. Ensure your responses are not verbose and use formatting that can be converted from text to speech."
},
{
"role": "user",
"content": "Crazy old Airbnb host turns lights and fan on and off flipping circuit breakers or some weird remote. Ban... | b4a2f575-500b-41af-979b-62de585f7f2d | {
"uuid": "b4a2f575-500b-41af-979b-62de585f7f2d",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide **minimal** responses with no additional information. **Avoid formatting** that cannot be spoken aloud."
},
{
"role": "user",
"content": "Role: You are a childrens story teller expert.\n\nTask: Ask me a series of questions to create a bespoke bedtime story... | a9ff5710-bf9e-434b-bd4f-7da343f1ca25 | {
"uuid": "a9ff5710-bf9e-434b-bd4f-7da343f1ca25",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "YOU MUST keep your responses as brief as possible while still answering the question. IMPORTANT: your output must be suitable for speech. YOU MUST NOT include any content that depends on visual layout."
},
{
"role": "user",
"content": "You know, while progress is ... | dcfedb7f-930e-4060-84ee-d461ad4045b5 | {
"uuid": "dcfedb7f-930e-4060-84ee-d461ad4045b5",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are speaking with the user on a phone call. Keep your responses brief and use only plain spoken language."
},
{
"role": "user",
"content": "Please provide a list (with corresponding time periods) of Ancient Greek cities that were the most influential cities fo... | e09e1e9e-9b3b-4a79-946f-a0aa9c514c06 | {
"uuid": "e09e1e9e-9b3b-4a79-946f-a0aa9c514c06",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Act as if you are on a phone call. Give the minimal answer needed to fully address the question using minimal formatting."
},
{
"role": "user",
"content": "血管内膜肉腫について詳細にまとめてみて\n特に転移がある進行肉腫の場合の化学療法についてをより詳細にお願い"
},
{
"role": "assistant",
"content": "血管内... | de6dc038-0f4c-44aa-a0d2-da1b7287574d | {
"uuid": "de6dc038-0f4c-44aa-a0d2-da1b7287574d",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Write your responses exactly how a person would say them out loud. Use only formatting that is easy to read as speech."
},
{
"role": "user",
"content": "After the railways and the electrification process, what was the next \"big impactful technology\" that changed... | eb346d26-3b7a-4d66-8f9d-4c558a27b00c | {
"uuid": "eb346d26-3b7a-4d66-8f9d-4c558a27b00c",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Please write your answers as if you are talking to someone in a real conversation. Your words will be read out loud by a computer voice, so make sure they sound natural and flow well. Try to use a conversational rhythm that feels like a human speaking. Avoid using complex sym... | 985f3d25-8757-4728-9521-7dbf9be58993 | {
"uuid": "985f3d25-8757-4728-9521-7dbf9be58993",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "YOU MUST keep your responses as brief as possible while still answering the question. IMPORTANT: your output must be suitable for speech. YOU MUST NOT include any content that depends on visual layout."
},
{
"role": "user",
"content": "you are an Unreal Engine exp... | 4f9305b0-3aca-45ae-b16c-13639249ba0e | {
"uuid": "4f9305b0-3aca-45ae-b16c-13639249ba0e",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide the minimal answer necessary to fully address the question, suitable for a spoken conversation."
},
{
"role": "user",
"content": "give me a summary of chapter 1 of the book making money moral by judith rodin and saadia madsbjerg"
},
{
"role": "assi... | 5ec3611b-ec84-4c9d-b5ee-420958b047d3 | {
"uuid": "5ec3611b-ec84-4c9d-b5ee-420958b047d3",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide a minimal answer while still addressing the question fully. Use only formatting that can be spoken aloud. Avoid all other structural elements."
},
{
"role": "user",
"content": "\"How life feels when you stop reading ML slop and start reading french theory ... | b01df3e9-fe23-4e94-a248-bb785b171441 | {
"uuid": "b01df3e9-fe23-4e94-a248-bb785b171441",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are on a phone call with the user. Use plain spoken language and be as brief as possible while still answering the question."
},
{
"role": "user",
"content": "questions to ask in a marketing interview to have them show you examples of marketing activities they... | a715321b-e779-46c3-b1ba-58ef7cbef7ee | {
"uuid": "a715321b-e779-46c3-b1ba-58ef7cbef7ee",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Write your responses exactly how a person would say them out loud. Use only formatting that is easy to read as speech."
},
{
"role": "user",
"content": "show me a data analysis learning plan"
},
{
"role": "assistant",
"content": "Alright, so if you're ... | bfc4c3ba-2154-424a-82b8-1c5dac75f622 | {
"uuid": "bfc4c3ba-2154-424a-82b8-1c5dac75f622",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Respond as a human speaking, providing the minimal answer that fully addresses the question. Ensure the formatting is suitable for text-to-speech conversion."
},
{
"role": "user",
"content": "If Bob has 3 brothers and a sister named Alice, how many brothers does A... | c51ca9c3-2449-4b24-8db9-d2c97dbd31f7 | {
"uuid": "c51ca9c3-2449-4b24-8db9-d2c97dbd31f7",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide the minimal answer needed to fully address the question, ensuring the text is easy to convert to speech."
},
{
"role": "user",
"content": "角色\n你是一位生肖组合对应专家，擅长解析各种生肖的性质及其之间的关系。\n\n技能\n技能1：生肖定位\n根据用户给出的详细描述，选出最贴近要求的三个吉美肖三个凶丑肖总共六个生肖。 例如： 吉美肖：兔龙蛇马羊鸡 凶丑肖：鼠牛虎猴狗猪... | 7ee25d7a-0488-4b56-ab3d-1623bb03d234 | {
"uuid": "7ee25d7a-0488-4b56-ab3d-1623bb03d234",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are communicating via voice comms in a video game. Keep responses short and to the point, and avoid any formatting that cannot be spoken aloud."
},
{
"role": "user",
"content": "Thomas is a paladin of freedom, sworn to the light. Isabella is secretly a vampire... | d8ad5313-48bf-4b3d-9f36-8f76c3358c38 | {
"uuid": "d8ad5313-48bf-4b3d-9f36-8f76c3358c38",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Your responses will be spoken aloud, so use minimal formatting and avoid writing a whole essay."
},
{
"role": "user",
"content": "Problème du nombre manquant :\n\nSoit un entier naturel N à 3 chiffres. On le multiplie par 7, puis par 11, puis par 13, dans cet ordr... | adfa991a-bda4-46c4-bf08-1ac8c3f2b224 | {
"uuid": "adfa991a-bda4-46c4-bf08-1ac8c3f2b224",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide responses that are as short and clear as possible, utilizing only speech-friendly formatting."
},
{
"role": "user",
"content": "Qué pesa más, medio kilo de plumas o un kilo de plomo?"
},
{
"role": "assistant",
"content": "Un kilo de plomo.",
... | 3548e49e-090a-4418-b9f3-3034b992c53f | {
"uuid": "3548e49e-090a-4418-b9f3-3034b992c53f",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Your responses will be processed by a text-to-speech system. Please keep your answers concise. Ensure the language is suitable for being spoken aloud."
},
{
"role": "user",
"content": "is 3.11 bigger than 3.2"
},
{
"role": "assistant",
"content": "No, ... | ed697000-6f0d-44fa-8357-276467d41dce | {
"uuid": "ed697000-6f0d-44fa-8357-276467d41dce",
"train_turns_mask": [
false,
false,
false,
false,
false,
false,
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "You are interacting with a user via **phone call or text message**. Please ensure your responses remain **brief** and concise. Use **only speech-friendly formatting** to ensure the text is easy to read or speak aloud."
},
{
"role": "user",
"content": "budget for p... | ef2a1e91-2fbd-45cb-a675-f3ca2642cc2f | {
"uuid": "ef2a1e91-2fbd-45cb-a675-f3ca2642cc2f",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Provide responses that are minimal and contain no additional information beyond the direct answer. Utilize only formatting that is suitable for oral delivery. Ensure the output remains concise and streamlined."
},
{
"role": "user",
"content": "Imagine 4 blocks on ... | d8e30eeb-8cee-48fe-a7a0-5e3b15906509 | {
"uuid": "d8e30eeb-8cee-48fe-a7a0-5e3b15906509",
"train_turns_mask": [
false,
false,
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
[
{
"role": "system",
"content": "Your responses are intended for a spoken conversation. Keep your contributions brief in nature. Use only formatting that feels natural when read aloud."
},
{
"role": "user",
"content": "What's a popular and well-maintained Python library for parsing YAML files?"
... | da18b86a-a496-4773-b39d-f5f0ec6fb745 | {
"uuid": "da18b86a-a496-4773-b39d-f5f0ec6fb745",
"train_turns_mask": [
false,
false,
true
]
} | [
"Super-3.5-VL"
] |
Nemotron-SFT-Instruction-Following-Chat-v4
Dataset Description
The Nemotron-SFT-Instruction-Following-Chat-v4 dataset is a collection of synthetic multi-turn conversations designed to strengthen models’ capabilities in general chat and instruction following. The data is generated using Kimi-K2.6, Gemma4-E4B-it and Qwen3.5-397B-A17B. It targets tasks where the model must respond to user requests in context and follow instructions across multiple turns.
The dataset combines three internally curated subsets: chat, multi-turn instruction following, and instruction following for response-length control. These subsets are similar in structure and goals to the nvidia/Nemotron-SFT-Instruction-Following-Chat-v3 dataset.
Chat
This subset contains synthetic multi-turn conversations for general chat. Seed user prompts are extended with simulated user turns and assistant responses, providing examples of answering follow-up requests in context. Records include separate assistant reasoning traces and masks that identify the turns to use for training.
Multi-Turn Instruction Following
This subset contains synthetic multi-turn conversations for instruction following. System instructions, user requests, and assistant responses provide examples of applying instructions across successive turns. Records include separate assistant reasoning traces and masks that identify the turns to use for training.
Instruction Following for Response Length Control
This subset contains synthetic multi-turn conversations with system instructions that control response length. User prompts and assistant responses provide examples of answering requests while following the specified length constraints. Records include separate assistant reasoning traces and masks that identify the turns to use for training.
This dataset is ready for commercial or non-commercial uses.
Dataset Owner(s)
NVIDIA Corporation
Dataset Creation Date
Created on: 10/09/2026
Last Modified on: 10/09/2026
Versioning:
Nemotron-SFT-Instruction-Following-Chat-v4
Previous Version(s):
- The
chatsubset replaces the chat data in Nemotron-SFT-Instruction-Following-Chat-v3. - The
instruction_following_for_response_length_controlandmulti_turn_instruction_followingsubsets supplement the instruction-following data in Nemotron-SFT-Instruction-Following-Chat-v2 and Nemotron-SFT-Instruction-Following-Chat-v3.
License/Terms of Use
Use of this dataset is governed by OpenMDW-1.1. Each user is responsible for checking the content of the dataset and the applicable licenses and determining if suitable for the intended use.
Additional Information: Includes prompts derived from WildChat-1M (Allen Institute for AI), licensed under the Open Data Commons Attribution License (ODC-By) v1.0 (https://huggingface.co/datasets/allenai/WildChat-1M). Includes prompts derived from LMArena Human Preference data (LMArena), user prompts licensed under the Creative Commons Attribution 4.0 International License (CC-BY-4.0) (https://huggingface.co/datasets/lmarena-ai/arena-human-preference-140k).
Intended Usage
This dataset is intended for researchers and developers training and evaluating models for general chat, multi-turn instruction following, and response-length control.
Use metadata.train_turns_mask to select the assistant turns for training. Keep the other messages as conversation context. For training without reasoning traces, remove reasoning_content from assistant messages while preserving their responses and the training mask.
chat
Use this subset to train models for general conversation and follow-up requests that depend on earlier turns. It replaces the chat data in Nemotron-SFT-Instruction-Following-Chat-v3. Train on the assistant turns selected by metadata.train_turns_mask.
multi_turn_instruction_following
Use this subset to train models to apply instructions across successive conversation turns. It supplements the instruction-following data in earlier Instruction-Following/Chat releases and can be blended with those subsets.
instruction_following_for_response_length_control
Use this subset to train models to follow response-length constraints while answering user requests across multiple turns. It can be blended with the instruction-following data in earlier Instruction-Following/Chat releases.
Dataset Characterization
Data Collection Method
Hybrid: Manually-Collected, Synthetic, Automated
Labeling Method
Hybrid: Manually-Labeled, Synthetic, Automated
Dataset Format
Modality: Text
Format: JSONL
Structure: Text + Metadata
Dataset Quantification
| Subset | Samples |
|---|---|
instruction_following_for_response_length_control |
266,190 |
chat |
914,716 |
multi_turn_instruction_following |
2,669,676 |
| Total | 3,850,582 |
Record count: 3,850,582 conversation records.
Total data storage: 80.74 GB.
Ethical Considerations
NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. Developers should work with their internal developer teams to ensure this dataset meets requirements for the relevant industry and use case and addresses unforeseen product misuse.
Please report quality, risk, security vulnerabilities or NVIDIA AI Concerns here.
- Downloads last month
- 113