Hugging Face's logo

Datasets:
p12315132
/
computational_limits_of_implicit_deductive_reasoning

Modalities:
Tabular
Text
Formats:
parquet
Size:
1M - 10M
Libraries:
Datasets
Dask
Polars
License:
Dataset card Data Studio Files
xet
 Community
2

Add task category and links to paper and code

#2
by nielsr HF Staff - opened
base: refs/heads/main
←
from: refs/pr/2
Discussion Files changed
+118
-112
Files changed (1) hide show
  1. README.md +118 -112
README.md CHANGED
@@ -1,112 +1,118 @@
1
- ---
2
- license: mit
3
- dataset_info:
4
- features:
5
- - name: preds
6
- sequence: string
7
- - name: rules
8
- sequence:
9
- - name: premises
10
- sequence: string
11
- - name: conclusion
12
- dtype: string
13
- - name: facts
14
- sequence: string
15
- - name: query
16
- dtype: string
17
- - name: label
18
- dtype: int64
19
- - name: depth
20
- dtype: int64
21
- splits:
22
- - name: train_lp_downsampled
23
- num_bytes: 659503714
24
- num_examples: 700000
25
- - name: train_rp_balanced_downsampled
26
- num_bytes: 850381210
27
- num_examples: 700000
28
- - name: validation_lp
29
- num_bytes: 6625773
30
- num_examples: 7000
31
- - name: validation_lp_balanced
32
- num_bytes: 7651259
33
- num_examples: 7000
34
- - name: validation_lp_balanced_deep_30_pred
35
- num_bytes: 15515151
36
- num_examples: 13000
37
- - name: validation_lp_balanced_deep_60_pred
38
- num_bytes: 27707171
39
- num_examples: 13000
40
- - name: validation_lp_star_balanced_deep_30_pred
41
- num_bytes: 16249806
42
- num_examples: 13000
43
- - name: validation_lp_star_balanced_deep_60_pred
44
- num_bytes: 25884528
45
- num_examples: 13000
46
- - name: validation_rp
47
- num_bytes: 7501908
48
- num_examples: 7000
49
- - name: validation_rp_balanced
50
- num_bytes: 8551772
51
- num_examples: 7000
52
- - name: validation_rp_balanced_deep_30_pred
53
- num_bytes: 19016725
54
- num_examples: 13000
55
- - name: validation_rp_balanced_deep_60_pred
56
- num_bytes: 32788646
57
- num_examples: 13000
58
- - name: validation_rp_balanced_3_premise
59
- num_bytes: 11934368
60
- num_examples: 7000
61
- - name: validation_rp_balanced_2_premise
62
- num_bytes: 8774321
63
- num_examples: 7000
64
- - name: validation_rp_balanced_1_premise
65
- num_bytes: 5474611
66
- num_examples: 7000
67
- - name: validation_rp_balanced_1_2_premise
68
- num_bytes: 6996474
69
- num_examples: 7000
70
- - name: validation_rp_balanced_2_3_premise
71
- num_bytes: 10312203
72
- num_examples: 7000
73
- download_size: 268682683
74
- dataset_size: 1720869640
75
- configs:
76
- - config_name: default
77
- data_files:
78
- - split: train_lp_downsampled
79
- path: data/train_lp_downsampled-*
80
- - split: train_rp_balanced_downsampled
81
- path: data/train_rp_balanced_downsampled-*
82
- - split: validation_lp
83
- path: data/validation_lp-*
84
- - split: validation_lp_balanced
85
- path: data/validation_lp_balanced-*
86
- - split: validation_lp_balanced_deep_30_pred
87
- path: data/validation_lp_balanced_deep_30_pred-*
88
- - split: validation_lp_balanced_deep_60_pred
89
- path: data/validation_lp_balanced_deep_60_pred-*
90
- - split: validation_lp_star_balanced_deep_30_pred
91
- path: data/validation_lp_star_balanced_deep_30_pred-*
92
- - split: validation_lp_star_balanced_deep_60_pred
93
- path: data/validation_lp_star_balanced_deep_60_pred-*
94
- - split: validation_rp
95
- path: data/validation_rp-*
96
- - split: validation_rp_balanced
97
- path: data/validation_rp_balanced-*
98
- - split: validation_rp_balanced_deep_30_pred
99
- path: data/validation_rp_balanced_deep_30_pred-*
100
- - split: validation_rp_balanced_deep_60_pred
101
- path: data/validation_rp_balanced_deep_60_pred-*
102
- - split: validation_rp_balanced_3_premise
103
- path: data/validation_rp_balanced_3_premise-*
104
- - split: validation_rp_balanced_2_premise
105
- path: data/validation_rp_balanced_2_premise-*
106
- - split: validation_rp_balanced_1_premise
107
- path: data/validation_rp_balanced_1_premise-*
108
- - split: validation_rp_balanced_1_2_premise
109
- path: data/validation_rp_balanced_1_2_premise-*
110
- - split: validation_rp_balanced_2_3_premise
111
- path: data/validation_rp_balanced_2_3_premise-*
112
- ---
 
 
 
 
 
 
 
1
+ ---
2
+ license: mit
3
+ task_categories:
4
+ - text-generation
5
+ dataset_info:
6
+ features:
7
+ - name: preds
8
+ sequence: string
9
+ - name: rules
10
+ sequence:
11
+ - name: premises
12
+ sequence: string
13
+ - name: conclusion
14
+ dtype: string
15
+ - name: facts
16
+ sequence: string
17
+ - name: query
18
+ dtype: string
19
+ - name: label
20
+ dtype: int64
21
+ - name: depth
22
+ dtype: int64
23
+ splits:
24
+ - name: train_lp_downsampled
25
+ num_bytes: 659503714
26
+ num_examples: 700000
27
+ - name: train_rp_balanced_downsampled
28
+ num_bytes: 850381210
29
+ num_examples: 700000
30
+ - name: validation_lp
31
+ num_bytes: 6625773
32
+ num_examples: 7000
33
+ - name: validation_lp_balanced
34
+ num_bytes: 7651259
35
+ num_examples: 7000
36
+ - name: validation_lp_balanced_deep_30_pred
37
+ num_bytes: 15515151
38
+ num_examples: 13000
39
+ - name: validation_lp_balanced_deep_60_pred
40
+ num_bytes: 27707171
41
+ num_examples: 13000
42
+ - name: validation_lp_star_balanced_deep_30_pred
43
+ num_bytes: 16249806
44
+ num_examples: 13000
45
+ - name: validation_lp_star_balanced_deep_60_pred
46
+ num_bytes: 25884528
47
+ num_examples: 13000
48
+ - name: validation_rp
49
+ num_bytes: 7501908
50
+ num_examples: 7000
51
+ - name: validation_rp_balanced
52
+ num_bytes: 8551772
53
+ num_examples: 7000
54
+ - name: validation_rp_balanced_deep_30_pred
55
+ num_bytes: 19016725
56
+ num_examples: 13000
57
+ - name: validation_rp_balanced_deep_60_pred
58
+ num_bytes: 32788646
59
+ num_examples: 13000
60
+ - name: validation_rp_balanced_3_premise
61
+ num_bytes: 11934368
62
+ num_examples: 7000
63
+ - name: validation_rp_balanced_2_premise
64
+ num_bytes: 8774321
65
+ num_examples: 7000
66
+ - name: validation_rp_balanced_1_premise
67
+ num_bytes: 5474611
68
+ num_examples: 7000
69
+ - name: validation_rp_balanced_1_2_premise
70
+ num_bytes: 6996474
71
+ num_examples: 7000
72
+ - name: validation_rp_balanced_2_3_premise
73
+ num_bytes: 10312203
74
+ num_examples: 7000
75
+ download_size: 268682683
76
+ dataset_size: 1720869640
77
+ configs:
78
+ - config_name: default
79
+ data_files:
80
+ - split: train_lp_downsampled
81
+ path: data/train_lp_downsampled-*
82
+ - split: train_rp_balanced_downsampled
83
+ path: data/train_rp_balanced_downsampled-*
84
+ - split: validation_lp
85
+ path: data/validation_lp-*
86
+ - split: validation_lp_balanced
87
+ path: data/validation_lp_balanced-*
88
+ - split: validation_lp_balanced_deep_30_pred
89
+ path: data/validation_lp_balanced_deep_30_pred-*
90
+ - split: validation_lp_balanced_deep_60_pred
91
+ path: data/validation_lp_balanced_deep_60_pred-*
92
+ - split: validation_lp_star_balanced_deep_30_pred
93
+ path: data/validation_lp_star_balanced_deep_30_pred-*
94
+ - split: validation_lp_star_balanced_deep_60_pred
95
+ path: data/validation_lp_star_balanced_deep_60_pred-*
96
+ - split: validation_rp
97
+ path: data/validation_rp-*
98
+ - split: validation_rp_balanced
99
+ path: data/validation_rp_balanced-*
100
+ - split: validation_rp_balanced_deep_30_pred
101
+ path: data/validation_rp_balanced_deep_30_pred-*
102
+ - split: validation_rp_balanced_deep_60_pred
103
+ path: data/validation_rp_balanced_deep_60_pred-*
104
+ - split: validation_rp_balanced_3_premise
105
+ path: data/validation_rp_balanced_3_premise-*
106
+ - split: validation_rp_balanced_2_premise
107
+ path: data/validation_rp_balanced_2_premise-*
108
+ - split: validation_rp_balanced_1_premise
109
+ path: data/validation_rp_balanced_1_premise-*
110
+ - split: validation_rp_balanced_1_2_premise
111
+ path: data/validation_rp_balanced_1_2_premise-*
112
+ - split: validation_rp_balanced_2_3_premise
113
+ path: data/validation_rp_balanced_2_3_premise-*
114
+ ---
115
+
116
+ This dataset accompanies the paper [The Scaling Properties of Implicit Deductive Reasoning in Transformers](https://huggingface.co/papers/2605.04330). It contains synthetic Horn clause reasoning problems with features such as rules, facts, queries, and labels, designed to study implicit deductive reasoning in depth-bounded Transformers.
117
+
118
+ Code for dataset generation, model training, and evaluation is available at: [https://github.com/envomp/Implicit-Deductive-Reasoning-in-Transformers](https://github.com/envomp/Implicit-Deductive-Reasoning-in-Transformers)