| --- |
| language: |
| - en |
| license: apache-2.0 |
| size_categories: |
| - 100K<n<1M |
| task_categories: |
| - text-generation |
| - question-answering |
| pretty_name: Math Reasoning SFT (100K) |
| tags: |
| - sft |
| - math |
| - reasoning |
| - mathematics |
| - step-by-step |
| - word-problems |
| - algebra |
| - geometry |
| - arithmetic |
| - probability |
| - statistics |
| - number-theory |
| - synthetic |
| configs: |
| - config_name: default |
| data_files: |
| - split: train |
| path: math-reasoning-sft-100k.jsonl |
| --- |
| |
| # Math Reasoning SFT (100K) |
|
|
| 100,000 math problems with detailed step-by-step solutions — ready for supervised fine-tuning of math reasoning models. |
|
|
| ## Dataset Description |
|
|
| **100,000 problems** across 8 mathematical categories and 3 difficulty levels: |
|
|
| ### Categories |
|
|
| | Category | Examples | Topics | |
| |---|---|---| |
| | `word_problems` | ~23,100 | Rate/time/distance, work problems, mixture, meeting/catch-up | |
| | `arithmetic` | ~15,400 | Percentages, profit/loss, ratios | |
| | `geometry` | ~15,400 | Area (rectangle, triangle, circle, trapezoid), Pythagorean theorem | |
| | `algebra` | ~15,200 | Linear equations, quadratic equations | |
| | `finance_math` | ~7,800 | Compound interest (annual/quarterly/monthly compounding) | |
| | `number_theory` | ~7,800 | GCD, LCM (Euclidean algorithm) | |
| | `probability` | ~7,700 | Dice, cards, balls, coins | |
| | `statistics` | ~7,600 | Mean, median, range | |
|
|
| ### Difficulty Distribution |
|
|
| | Level | Examples | Description | |
| |---|---|---| |
| | `easy` | ~48,000 | Single-step or direct formula application | |
| | `medium` | ~44,300 | Multi-step problems requiring setup | |
| | `hard` | ~7,700 | Problems requiring simultaneous equations or complex setup | |
|
|
| ## Format |
|
|
| ```json |
| { |
| "problem": "A car travels at 60 km/h for 2.5 hours. How far does it travel?", |
| "solution": "Step 1: Use the formula: Distance = Speed × Time\n Distance = 60 km/h × 2.5 h = 150 km\n\nAnswer: 150 km", |
| "answer": "150 km", |
| "category": "word_problems", |
| "difficulty": "easy", |
| "id": "abc123" |
| } |
| ``` |
|
|
| ## Solution Structure |
|
|
| Every solution follows a consistent multi-step format: |
| 1. **Formula identification** — states the relevant formula or approach |
| 2. **Substitution** — plugs in the specific numbers |
| 3. **Step-by-step calculation** — shows intermediate results |
| 4. **Final answer** — clearly labeled `Answer:` line |
|
|
| ## Use Case |
|
|
| - SFT for math reasoning capability in LLMs |
| - Training step-by-step problem solving behavior |
| - Curriculum learning: filter by `difficulty` for progressive training |
| - Evaluation: held-out test sets for math benchmarks |
| - Chain-of-thought distillation |
|
|
| ## Filtering Examples |
|
|
| ```python |
| import json |
| |
| # Load only medium/hard problems |
| with open("math-reasoning-sft-100k.jsonl") as f: |
| data = [json.loads(l) for l in f] |
| |
| hard_algebra = [r for r in data if r["difficulty"] in ("medium","hard") and r["category"] == "algebra"] |
| ``` |
|
|
| ## License |
|
|
| Apache 2.0 |
|
|
