Datasets:
Update README.md
Browse files
README.md
CHANGED
@@ -1,3 +1,74 @@
|
|
1 |
-
---
|
2 |
-
license: cc-by-4.0
|
3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
license: cc-by-4.0
|
3 |
+
task_categories:
|
4 |
+
- text-generation
|
5 |
+
language:
|
6 |
+
- en
|
7 |
+
tags:
|
8 |
+
- logic
|
9 |
+
- syllogisms
|
10 |
+
- premise_selection
|
11 |
+
- synthetic
|
12 |
+
size_categories:
|
13 |
+
- 100K<n<1M
|
14 |
+
---
|
15 |
+
|
16 |
+
# Dataset Summary
|
17 |
+
|
18 |
+
**Syllogistic-logic** is a synthetic dataset designed to evaluate the logical reasoning abilities of LLMs. It focuses on the task of *logical premise selection* — identifying the **minimal** set of premises in a knowledge base that entails a given hypothesis. The dataset is built on the syllogistic fragment of first-order logic and supports systematic generalization experiments, including generalization to unseen knowledge bases and reasoning with longer or shorter inference chains than seen during training.
|
19 |
+
|
20 |
+
The dataset is used to train and evaluate models using the **MIND** (**M**eta-learning for **IN**-context **D**eduction) fine-tuning approach, which adapts few-shot meta-learning to logical reasoning. Each row in the dataset includes a knowledge base, study examples (support), a query hypothesis, and the set of premises from which the query hypothesis can be derived.
|
21 |
+
|
22 |
+
Further details can be found in the paper where the dataset was proposed: [A MIND for Reasoning: Meta-learning for In-context Deduction](https://arxiv.org/abs/2505.14313)
|
23 |
+
|
24 |
+
## Features
|
25 |
+
|
26 |
+
Each example in the dataset includes the following fields:
|
27 |
+
|
28 |
+
| Column | Type | Description |
|
29 |
+
|-------------------------|----------|------------------------------------------------------------------------------------------------------|
|
30 |
+
| `knowledge_base` | `string` | Concatenated premises representing the KB. |
|
31 |
+
| `study_examples_all` | `string` | Concatenated study (support) examples for core generalization (see paper). |
|
32 |
+
| `study_examples_comp` | `string` | Concatenated study (support) examples for long->short generalization (see paper). |
|
33 |
+
| `study_examples_rec` | `string` | Concatenated study (support) examples for short->long generalization (see paper). |
|
34 |
+
| `query_hyp` | `string` | Query hypothesis to be proven. |
|
35 |
+
| `query_inf` | `string` | Minimal set of premises within KB that entail the query hypothesis. |
|
36 |
+
| `kb_id` | `string` | Identifier of the knowledge base. |
|
37 |
+
| `pword_a` | `int64` | Pseudowords assignment. |
|
38 |
+
| `kb_permut` | `int64` | Permutation index of the orderd of premises within KB. |
|
39 |
+
| `#prem_kb` | `int64` | Total premises in the KB. |
|
40 |
+
| `inf_length` | `int64` | Number of minimal premises necessary to derive the query hypotheisis. |
|
41 |
+
| `inf_type` | `int64` | Type of syllogistic inference (from 1 to 7). |
|
42 |
+
|
43 |
+
|
44 |
+
## To load the dataset, run:
|
45 |
+
|
46 |
+
```python
|
47 |
+
from datasets import load_dataset
|
48 |
+
|
49 |
+
data_files = {
|
50 |
+
"train": "train.csv",
|
51 |
+
"validation": "validation.csv",
|
52 |
+
"test": "test.csv",
|
53 |
+
"test_ood_constants": "test_ood_constants.csv",
|
54 |
+
"test_ood_support": "test_ood_support.csv",
|
55 |
+
"test_ood_words": "test_ood_words.csv",
|
56 |
+
}
|
57 |
+
full_data = load_dataset("leobertolazzi/syllogistic-logic", data_files=data_files)
|
58 |
+
```
|
59 |
+
|
60 |
+
## Licensing
|
61 |
+
This dataset is licensed under a [CC BY-SA 4.0 License](https://creativecommons.org/licenses/by-sa/4.0/).
|
62 |
+
|
63 |
+
## Cite
|
64 |
+
```bibtex
|
65 |
+
@misc{bertolazzi-et-al-2025-mind,
|
66 |
+
title={A MIND for Reasoning: Meta-learning for In-context Deduction},
|
67 |
+
author={Leonardo Bertolazzi and Manuel Vargas Guzmán and Raffaella Bernardi and Maciej Malicki and Jakub Szymanik},
|
68 |
+
year={2025},
|
69 |
+
eprint={2505.14313},
|
70 |
+
archivePrefix={arXiv},
|
71 |
+
primaryClass={cs.CL},
|
72 |
+
url={https://arxiv.org/abs/2505.14313},
|
73 |
+
}
|
74 |
+
```
|