Datasets:
File size: 4,835 Bytes
e4a0376 0268b9b e4a0376 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 | ---
license: other
task_categories:
- text-generation
tags:
- k2-horizon
- training-data
- parquet
- math
- reasoning
configs:
- config_name: math-thinking-qwen
data_files:
- split: train
path: "math-thinking-qwen/*.parquet"
- config_name: math-thinking-oss
data_files:
- split: train
path: "math-thinking-oss/*.parquet"
- config_name: math-rewrite
data_files:
- split: train
path: "math-rewrite/*.parquet"
- config_name: math-dialogue
data_files:
- split: train
path: "math-dialogue/*.parquet"
- config_name: agentic-math-dialogue
data_files:
- split: train
path: "agentic-math-dialogue/*.parquet"
---
# Math-Reasoning
## Dataset Description
Mathematical problem-solving, rewriting, and dialogue data for reasoning-oriented language-model training. This repository is part of the [K2 Horizon collection](https://huggingface.co/collections/IFM/k2-horizon).
The release is organized as one Hugging Face configuration per subset. Every configuration has a `train` split backed by Parquet shards, which supports Dataset Viewer inspection and streaming access.
## K2 Horizon Dataset Series
| Dataset repository | Focus | Configurations |
| --- | --- | ---: |
| [IFM/TxT360-v2](https://huggingface.co/datasets/IFM/TxT360-v2) | Web and question-answering text | 3 |
| [IFM/Code-Reasoning](https://huggingface.co/datasets/IFM/Code-Reasoning) | Code reasoning and task synthesis | 7 |
| [IFM/Math-Reasoning](https://huggingface.co/datasets/IFM/Math-Reasoning) | Mathematical reasoning and dialogue | 5 |
| [IFM/SFT-Reasoning](https://huggingface.co/datasets/IFM/SFT-Reasoning) | Instruction following and supervised fine-tuning | 2 |
| [IFM/Pretrain-Behaviors](https://huggingface.co/datasets/IFM/Pretrain-Behaviors) | Behavior-focused continued-pretraining data | 7 |
## Dataset Configurations
| Configuration | Catalog source | Data files |
| --- | --- | --- |
| `math-thinking-qwen` | `math-qwen-no-think` | `math-thinking-qwen/*.parquet` |
| `math-thinking-oss` | `math-oss-cleaned` | `math-thinking-oss/*.parquet` |
| `math-rewrite` | `math-rewrite` | `math-rewrite/*.parquet` |
| `math-dialogue` | `math-dialogue` | `math-dialogue/*.parquet` |
| `agentic-math-dialogue` | `agentic-math-dialogue-cleaned` | `agentic-math-dialogue/*.parquet` |
## Repository Structure
```text
README.md
math-thinking-qwen/
<source-file>-<stable-id>-00000.parquet
<source-file>-<stable-id>-00001.parquet
math-thinking-oss/
<source-file>-<stable-id>-00000.parquet
<source-file>-<stable-id>-00001.parquet
math-rewrite/
<source-file>-<stable-id>-00000.parquet
<source-file>-<stable-id>-00001.parquet
math-dialogue/
<source-file>-<stable-id>-00000.parquet
<source-file>-<stable-id>-00001.parquet
agentic-math-dialogue/
<source-file>-<stable-id>-00000.parquet
<source-file>-<stable-id>-00001.parquet
```
The shard prefix is derived from the source JSONL filename and a stable identifier. Updating one source JSONL file replaces only that file's Parquet shards.
## Data Fields
Records originate as JSON objects and are converted to Parquet for release. Field names and nested structures can differ by configuration. Inspect `features` before building a processing pipeline:
```python
from datasets import load_dataset
dataset = load_dataset(
"IFM/Math-Reasoning",
"math-thinking-qwen",
split="train",
streaming=True,
)
print(dataset.features)
print(next(iter(dataset)))
```
## Data Provenance and Processing
The configurations in this repository are selected from the data inventory used to prepare the K2 Horizon training mixture. Only release-approved configurations are included. JSONL records are converted to Parquet without intentionally renaming application-level fields.
Individual configurations may have undergone source-specific filtering, cleaning, deduplication, quality scoring, or synthetic-data generation. Users should evaluate each configuration for their target use case and inspect the available provenance metadata.
## Intended Use
This dataset is intended for language-model training and research. The configurations can be streamed independently, combined with user-defined sampling weights, or inspected through the Hugging Face Dataset Viewer.
## Limitations and Responsible Use
Large-scale training data can contain factual errors, duplicated material, sensitive topics, stereotypes, unsafe content, and other artifacts. Dataset users are responsible for performing evaluations, risk assessment, and filtering appropriate to their application.
## License and Terms of Use
This repository contains multiple configurations that may have different source terms. Users are responsible for reviewing the applicable provenance and license information for the configurations they use and for determining suitability for their intended purpose.
|