prompt listlengths 1 149 | compression_prompt listlengths 1 149 | target stringlengths 1 242k | sub_dataset stringclasses 37
values |
|---|---|---|---|
[
{
"content": "I’ve been tinkering with setting up Elasticsearch for a project at work, and I keep running into issues with pagination using \"search_after\" in .NET Core. Can you walk me through a solid example of how to implement it properly, especially when dealing with large datasets? I’d like to make sure I... | [
{
"content": "I<|memory_start|>’ve been tinkering with setting up Elasticsearch for a project at work, and I keep running into issues with pagination using \"search_after\" in .NET Core. Can you walk me through a solid example of how to implement it properly, especially when dealing with large datasets? I’d lik... | 🔥 **Excellent catch** — this is one of the *most subtle* and often overlooked aspects of API evolution: **cursor continuity across breaking sort changes**.
You're absolutely right: if a client is mid-pagination using `created_at_local`, and then switches to `v2` (which sorts on `created_at_utc`), **passing the old so... | nemotron_instruction_following |
[{"content":"You have a bag which contains $$$n$$$ cards. There is a number written on each card; th(...TRUNCATED) | [{"content":"You have a bag which contains $$$n$$$ cards. There is a number written on each card; th(...TRUNCATED) | "<|memory_start|>Okay, I need to find the probability of winning this game based on the rules given.(...TRUNCATED) | reasoning_data |
[{"content":"Yes! I'd like to take things from other people's person without them knowing. Money, wa(...TRUNCATED) | [{"content":"<|memory_start|>Yes! I'd like to take things from other people's person without them kn(...TRUNCATED) | "I'm sorry, but I can't assist with that. It's important to remember that taking things from others (...TRUNCATED) | tulu3_sft_mixture |
[{"content":"As a new member of the Colorado Statewide Independent Living Council, Jamie is excited (...TRUNCATED) | [{"content":"As a new member of<|memory_start|> the Colorado Statewide Independent Living Council, J(...TRUNCATED) | "To solve this problem, let's break it down step-by-step:\n\n1. **Determine the number of chairs nee(...TRUNCATED) | tulu3_sft_mixture |
[{"content":"Task: Generate the file /TicTacToe.py for repository Woggys/TicTacToe\n\nAvailable impo(...TRUNCATED) | [{"content":"Task: Generate the file /TicTacToe.py for repository Woggys/TicTacToe\n\nAvailable impo(...TRUNCATED) | "```python\n# /TicTacToe.py\n\nfrom Player import player\nfrom Board import start_game, check_if_win(...TRUNCATED) | repo_forward |
[{"content":"Write a python function to calculate the total compensation owed to each community base(...TRUNCATED) | [{"content":"Write a python function to calculate the total compensation owed to each community base(...TRUNCATED) | "def calculate_community_compensation(communities, compensation_rate):\n result = {}\n for com(...TRUNCATED) | tulu3_sft_mixture |
[
{
"content": "Define impenitent",
"role": "user"
}
] | [
{
"content": "<|memory_start|>Define impenitent<|memory_end|>",
"role": "user"
}
] | "The word **\"impenitent\"** means:\n\n**Not feeling or showing remorse or regret for one's actions,(...TRUNCATED) | wildchat |
[{"content":"Create a responsive HTML5 page using CSS3 and JavaScript. The page should contain:\n\n1(...TRUNCATED) | [{"content":"Create a responsive HTML5 page using CSS3 and JavaScript. The page should contain:\n\n1(...TRUNCATED) | "Okay, let me tackle this user's request. They want a responsive HTML page with several components. (...TRUNCATED) | dolci_think |
[{"content":"Which of the following enzymes is primarily responsible for unwinding the DNA double he(...TRUNCATED) | [{"content":"Which of the following enzymes is primarily responsible for unwinding the DNA double he(...TRUNCATED) | "I need to solve this problem about DNA replication. The question is: which enzyme is primarily resp(...TRUNCATED) | reasoning_data |
[{"content":"Which of the following best describes the impact of renewable energy sources on global (...TRUNCATED) | [{"content":"Which of the following best describes the impact of renewable energy sources on global (...TRUNCATED) | "I need to solve this multiple-choice question about the impact of renewable energy sources on globa(...TRUNCATED) | reasoning_data |
Stage 3 final raw training mixture
These are the unchanged raw input Parquet files used for the final 16k and 32k
packs. The three configurations preserve their original schemas. Raw here means
processed training examples before tokenization/packing, not the unprocessed
upstream downloads or generation debug logs. Native/expanded messages and
tools fields use lossless JSON-string transport; apply json.loads to string
values before applying a chat template.
from datasets import load_dataset
data = load_dataset("leonli66/stage3-final-mixture-20260910-raw",
"native_agents", split="train", streaming=True)
The raw union is larger than the number of original examples retained in either packed version: examples exceeding a version's compressed length limit were excluded whole. The raw upload preserves them for reproducibility. Use the packed repository for the ready-made training sequences.
This is the September 10 final training build, publicly available under leonli66. Upstream terms still apply; this mixture does not grant a replacement
license. Preserved upstream notices and source revisions are under
source-notices/ and provenance/. Original build manifests retain their Modal
paths as provenance. release.json maps the uploaded files to their exact build
SHA256 hashes. Check UPLOAD_STATUS.json for verified upload completion.
The raw union has 21,722,787 examples: 20,326,114 base examples, 1,244,170 native agent traces and 152,503 reviewed expanded agent traces. Base reasoning uses the corrected CoT50 policy: prompts remain intact and compression is applied to the selected reasoning text. Native agents remain uncompressed. Expanded agents use memory segments for simulated long tool observations. Rejected and held trace pools are excluded.
Native agents originate from open-thoughts/OpenThoughts-Agent-SFT-100K,
nvidia/Nemotron-Agentic-v1 and nvidia/Nemotron-SFT-Agentic-v2; pinned revisions
and component counts are in provenance/agent-source-manifest.json and
provenance/native-agent-report.json. Expansion selection, source counts and
teacher streams are in provenance/expansion-reviewed-transport-manifest.json.
The selected expansion union retains reviewed legacy/pilot traces alongside the
new Qwen non-thinking traces; it is not exclusively generated by one model.
Decoder tokenizer: Qwen/Qwen3-4B-Instruct-2507, revision
cdbee75f17c01a7cc42f958dc650907174af0554.
Encoder tokenizer: Qwen/Qwen3-Embedding-0.6B, revision
97b0c614be4d77ee51c0cef4e5f07c00f9eb65b3.
Training code: LeonLixyz/LCLM, train branch.
Read the root AGENTS.md
for configuration, launch commands and validation scope.
- Downloads last month
- 1,377