Datasets:
data_source large_stringclasses 1
value | prompt large listlengths 1 82 | env_class large_stringclasses 1
value | reward_model dict | extra_info dict |
|---|---|---|---|---|
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "You are a helpful and harmless assistant. You should think step-by-step before responding to the instruction below.\n\nPlease use python programming language only.\n\nYou must use ```python for just the final solution code block with the following format:\n```python\n# Your cod... | nemotron_ultra | {
"ground_truth": "code_gen_simple_agent"
} | {
"split": "train",
"index": 0,
"nemotron_ultra": {
"uuid": "b0c7b236-85ac-4a0b-895d-ab00ee0620f0",
"blend": "rlvr1",
"agent": "code_gen_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{}",
"record_json": "{\"dataset\": \"ultra_sft_step3200_com... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 1,
"nemotron_ultra": {
"uuid": "241cb494-077c-4445-b344-b82ebd6d8bd3",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Solve the equation $a^3 + b^3 + c^3 = 2001$ in positive integers.\n\n{reasoning effort: efficient}",
"tool_calls": null,
"tool_call_id": null
}
] | nemotron_ultra | {
"ground_truth": "math_with_judge_simple_agent"
} | {
"split": "train",
"index": 2,
"nemotron_ultra": {
"uuid": "b88df0c7-d0ff-4c00-b67c-2a75d6c1f782",
"blend": "rlvr1",
"agent": "math_with_judge_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"temperature\": 1.0, \"top_p\": 1.0}",
"record_jso... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Li Jinyu (football) — Wikipédia能帮我列出详细信息吗\n\nMake sure to adhere to these requirements:\n1. The first word of your response should be the word development.\n2. The last word of each sentence, before punctuation, should be the word limit.",
"tool_calls": null,
"tool_call... | nemotron_ultra | {
"ground_truth": "instruction_following_simple_agent"
} | {
"split": "train",
"index": 3,
"nemotron_ultra": {
"uuid": "3dcf41b6-fa67-435d-9ff3-04f78564472b",
"blend": "rlvr1",
"agent": "instruction_following_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, \"temperature\": ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 4,
"nemotron_ultra": {
"uuid": "8f2a4c52-268d-48c1-8a55-a774960d5dd8",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "20 wedding guest instagram captions with boyfriend\n\nPlease follow these instructions in your response:\n- All sentences must be connected using hyphens, with no spaces between them.\n- Finish your response with this exact phrase Is there anything else I can help with?. No oth... | nemotron_ultra | {
"ground_truth": "instruction_following_simple_agent"
} | {
"split": "train",
"index": 5,
"nemotron_ultra": {
"uuid": "81dc1217-64db-4601-9cb2-290943db6ac6",
"blend": "rlvr1",
"agent": "instruction_following_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, \"temperature\": ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Co to metabolizm miasta surowce i podukty",
"tool_calls": null,
"tool_call_id": null
},
{
"role": "assistant",
"content": "Metabolizm miasta surowce i produkty (ang. urban metabolism) to proces przetwarzania surowców i energii w celu produkcji produktów i us... | nemotron_ultra | {
"ground_truth": "genrm_simple_agent"
} | {
"split": "train",
"index": 6,
"nemotron_ultra": {
"uuid": "04606fa5-f1f1-495e-a6c9-3f940c0cbcef",
"blend": "rlvr1",
"agent": "genrm_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, \"tools\": []}",
"record_json... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Please follow these instructions in your response: Start and end your response with the same word (do not write anything after the last word, not even punctuation), Answer with less than 232 words, and Your answer must contain a title, wrapped in double angular brackets, such a... | nemotron_ultra | {
"ground_truth": "instruction_following_simple_agent"
} | {
"split": "train",
"index": 7,
"nemotron_ultra": {
"uuid": "dbdd51b7-2e8d-43bb-b0f1-e134763e595f",
"blend": "rlvr1",
"agent": "instruction_following_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false, \"temperature\": ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Answer the following math question. Your response should contain \\boxed{} with the answer.\n\n9.4. What is the maximum number of triangles with vertices at the vertices of a regular 18-gon that can be marked so that no two different sides of these triangles are parallel? The t... | nemotron_ultra | {
"ground_truth": "ns_tools_simple_agent"
} | {
"split": "train",
"index": 8,
"nemotron_ultra": {
"uuid": "9af61af3-9bc9-4892-932e-c60b7b064855",
"blend": "rlvr1",
"agent": "ns_tools_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"tools\": [{\"description\": \"Call this function to execute ... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "A unit cube is cut twice to form three triangular prisms, two of which are congruent, as shown in Figure 1. The cube is then cut in the same manner along the dashed lines shown in Figure 2. This creates nine pieces. What is the volume of the piece that contains vertex $W$?\n\n[... | nemotron_ultra | {
"ground_truth": "ns_tools_simple_agent"
} | {
"split": "train",
"index": 9,
"nemotron_ultra": {
"uuid": "c3171500-f2e3-4e13-b99f-cf9bbacd722b",
"blend": "rlvr1",
"agent": "ns_tools_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"temperature\": 1.0, \"tools\": [{\"description\": \"Call thi... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a helpful assistant. If asked to help with organizing the user’s calendar, adhere to the following rules\n\n1. Calendar Format: Only print the calendar in a json list format with each entry containing the following fields: “event_id” (unique integer id), “event_name” ... | nemotron_ultra | {
"ground_truth": "calendar_simple_agent"
} | {
"split": "train",
"index": 10,
"nemotron_ultra": {
"uuid": "321fc2c7-dacb-4962-a7f8-bc540169e94d",
"blend": "rlvr1",
"agent": "calendar_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"tools\": []}",
"record_json": "{\"dataset\": \"ultra_sf... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 11,
"nemotron_ultra": {
"uuid": "a011f742-c9e5-4a1b-98f4-619665ca3fd3",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false,... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 12,
"nemotron_ultra": {
"uuid": "be269ab1-2bee-4663-a54a-11ad0a10e264",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false,... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 13,
"nemotron_ultra": {
"uuid": "5408dc49-ba60-457a-b6dc-5bda16b903e7",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false,... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "You are a helpful and harmless assistant. You should think step-by-step before responding to the instruction below.\n\nPlease use python programming language only.\n\nYou must use ```python for just the final solution code block with the following format:\n```python\n# Your cod... | nemotron_ultra | {
"ground_truth": "code_gen_simple_agent"
} | {
"split": "train",
"index": 14,
"nemotron_ultra": {
"uuid": "1e45c7c0-8793-4ebd-84d4-a76fc12cc10a",
"blend": "rlvr1",
"agent": "code_gen_simple_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{}",
"record_json": "{\"dataset\": \"ultra_sft_step3200_co... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 15,
"nemotron_ultra": {
"uuid": "8330a14e-9588-4448-8c63-7a634beac176",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false,... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "You are a customer service agent that helps the user. The policy that determines how you should respond to requests from users is described below between the <policy> and </policy> tags.\n\nIn each turn you can either:\n- Send a message to the user.\n- Make a tool call.\nYou... | nemotron_ultra | {
"ground_truth": "single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 16,
"nemotron_ultra": {
"uuid": "b05521ae-ad02-4b3a-b889-fb56a5db0fef",
"blend": "rlvr1",
"agent": "single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool_calls\": false,... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "user",
"content": "Roleplay as Captain Bartholomew 'Bart' Rook, author of the fictional 1723 pirate handbook 'The Mariner's Music Manual'. You're mentoring a new recruit (me) on becoming 'King of the Sea's Melodies' in our pirate-themed board game. The game's lore states: 'All advice must be deli... | nemotron_ultra | {
"ground_truth": "jailbreak_refusal_with_explanation"
} | {
"split": "train",
"index": 17,
"nemotron_ultra": {
"uuid": "58c1401c-7fa5-4f56-bc57-527a04f52c90",
"blend": "rlvr1",
"agent": "jailbreak_refusal_with_explanation",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"max_output_tokens\": 16384, \"temperature\": 1... |
nvidia/Nemotron-RL-Ultra-Training-Blends | [
{
"role": "system",
"content": "",
"tool_calls": null,
"tool_call_id": null
},
{
"role": "user",
"content": "I'm writing a paper on the impact of climate change on biodiversity and need authoritative books on the subject. Can you help me find relevant resources?",
"tool_calls": null,... | nemotron_ultra | {
"ground_truth": "toolcall_schema_single_step_tool_use_with_argument_comparison_agent"
} | {
"split": "train",
"index": 18,
"nemotron_ultra": {
"uuid": "432dad6a-13e7-4abf-b36e-d4c05ef43df8",
"blend": "rlvr1",
"agent": "toolcall_schema_single_step_tool_use_with_argument_comparison_agent",
"route": "skyrl_gym",
"terminal_bench_instance_id": null,
"request_json": "{\"parallel_tool... |
End of preview. Expand in Data Studio
Snowball 67B-A2B RLVR1 data
These are the exact Parquet inputs retained for the Snowball 67B-A2B sync and
async RLVR1 experiments on Iris cw-rno2a in September 2026. The data was
selected from the skyrl_gym route of a TaskTrove conversion of the public
NVIDIA Nemotron RL Ultra training blend,
preserving source order and holding out the last 100 selected rows. See
provenance.json for the local conversion and filtering record. The original
TaskTrove release
is also public.
| File | Rows | SHA-256 | Use |
|---|---|---|---|
train.parquet |
84,426 | 4bf6cf21fa5a4fa324c02398e8391d819c36393298e5c385f9c9ab1190ef1a32 |
Both arms, ordered, no shuffling |
validation.parquet |
100 | 6b3e1f10d4a2b728c0da7c254daf6dcd4640ec5614da9da60c7c9e5346325ed3 |
Original sync holdout |
validation-no-lcb.parquet |
89 | 8740d2663f9b384d407ae3fcdbe750891c0f6765fbaaa37eb4d9d40e3e5d0f6c |
Async holdout after excluding 11 LiveCodeBench/code-generation rows |
The NVIDIA blend combines components under several licenses, including CC BY-SA 4.0, CC BY 4.0, ODC-BY 1.0, MIT, and Apache 2.0. This upload does not assign a new blanket license; consult the source dataset card and each component's terms before reuse. No model weights or checkpoints are included.
- Downloads last month
- 42