Add README
Browse files
README.md
CHANGED
|
@@ -2,46 +2,20 @@
|
|
| 2 |
license: gpl-3.0
|
| 3 |
pretty_name: Calculator
|
| 4 |
size_categories:
|
| 5 |
-
- 10K<n<100K
|
| 6 |
task_categories:
|
| 7 |
-
- text-generation
|
| 8 |
language:
|
| 9 |
-
- en
|
| 10 |
configs:
|
| 11 |
-
- config_name: default
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
dataset_info:
|
| 20 |
-
features:
|
| 21 |
-
- name: prompt
|
| 22 |
-
dtype: string
|
| 23 |
-
- name: completion
|
| 24 |
-
dtype: string
|
| 25 |
-
- name: answer
|
| 26 |
-
dtype: string
|
| 27 |
-
- name: expression
|
| 28 |
-
dtype: string
|
| 29 |
-
- name: steps
|
| 30 |
-
dtype: int64
|
| 31 |
-
- name: text
|
| 32 |
-
dtype: string
|
| 33 |
-
splits:
|
| 34 |
-
- name: train
|
| 35 |
-
num_bytes: 11115287
|
| 36 |
-
num_examples: 15000
|
| 37 |
-
- name: validation
|
| 38 |
-
num_bytes: 557003
|
| 39 |
-
num_examples: 750
|
| 40 |
-
- name: test
|
| 41 |
-
num_bytes: 2224087
|
| 42 |
-
num_examples: 3000
|
| 43 |
-
download_size: 6250337
|
| 44 |
-
dataset_size: 13896377
|
| 45 |
---
|
| 46 |
|
| 47 |
# Arithmetic Expression Dataset Generator
|
|
@@ -54,7 +28,11 @@ This is a synthetic deterministic dataset with nested arithmetic expressions and
|
|
| 54 |
- square brackets: `[]`
|
| 55 |
- curly braces: `{}`
|
| 56 |
|
| 57 |
-
and operators:
|
|
|
|
|
|
|
|
|
|
|
|
|
| 58 |
|
| 59 |
The dataset is balanced across reduction depths from 1 to 15 and is intended for SLMs that learn structured mathematical reasoning from examples.
|
| 60 |
|
|
@@ -81,32 +59,20 @@ Each step from 1 to 15 appears exactly:
|
|
| 81 |
- 50 times in validation
|
| 82 |
- 200 times in test
|
| 83 |
|
| 84 |
-
The symbol-length distribution:
|
| 85 |
-
|
| 86 |
-

|
| 87 |
-
|
| 88 |
## Example
|
| 89 |
|
| 90 |
```text
|
| 91 |
### Expression
|
| 92 |
-
Calculate:
|
| 93 |
|
| 94 |
### Answer
|
| 95 |
<think>
|
| 96 |
-
Start:
|
| 97 |
-
1. (
|
| 98 |
-
2. (
|
| 99 |
-
3. (
|
| 100 |
-
4. (12 * 19) = 228
|
| 101 |
-
5. (36 - 228) = -192
|
| 102 |
-
6. (8 + -192) = -184
|
| 103 |
-
7. (15 * 13) = 195
|
| 104 |
-
8. (195 - 12) = 183
|
| 105 |
-
9. (-184 - 183) = -367
|
| 106 |
-
10. (-367 + 11) = -356
|
| 107 |
-
11. (8 * -356) = -2848
|
| 108 |
</think>
|
| 109 |
-
<answer>
|
| 110 |
```
|
| 111 |
|
| 112 |
## Dataset generation
|
|
|
|
| 2 |
license: gpl-3.0
|
| 3 |
pretty_name: Calculator
|
| 4 |
size_categories:
|
| 5 |
+
- 10K<n<100K
|
| 6 |
task_categories:
|
| 7 |
+
- text-generation
|
| 8 |
language:
|
| 9 |
+
- en
|
| 10 |
configs:
|
| 11 |
+
- config_name: default
|
| 12 |
+
data_files:
|
| 13 |
+
- split: train
|
| 14 |
+
path: data/train-*
|
| 15 |
+
- split: validation
|
| 16 |
+
path: data/validation-*
|
| 17 |
+
- split: test
|
| 18 |
+
path: data/test-*
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 19 |
---
|
| 20 |
|
| 21 |
# Arithmetic Expression Dataset Generator
|
|
|
|
| 28 |
- square brackets: `[]`
|
| 29 |
- curly braces: `{}`
|
| 30 |
|
| 31 |
+
and operators:
|
| 32 |
+
|
| 33 |
+
- `+`
|
| 34 |
+
- `-`
|
| 35 |
+
- `*`
|
| 36 |
|
| 37 |
The dataset is balanced across reduction depths from 1 to 15 and is intended for SLMs that learn structured mathematical reasoning from examples.
|
| 38 |
|
|
|
|
| 59 |
- 50 times in validation
|
| 60 |
- 200 times in test
|
| 61 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 62 |
## Example
|
| 63 |
|
| 64 |
```text
|
| 65 |
### Expression
|
| 66 |
+
Calculate: (11 + (296 - 98))
|
| 67 |
|
| 68 |
### Answer
|
| 69 |
<think>
|
| 70 |
+
Start: (11 + (296 - 98))
|
| 71 |
+
1. (74 * 4) = 296
|
| 72 |
+
2. (296 - 98) = 198
|
| 73 |
+
3. (11 + 198) = 209
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 74 |
</think>
|
| 75 |
+
<answer>209</answer>
|
| 76 |
```
|
| 77 |
|
| 78 |
## Dataset generation
|