Hugging Face's logo

deepset
/
roberta-base-squad2

Question Answering
Transformers
PyTorch
TensorFlow
JAX
Rust
Safetensors
English
roberta
Eval Results
Model card Files
xet
Community
33
fatima113 commited on
Commit
35cafe7
路
verified 路
1 Parent(s): adc3b06

add AIBOM

Browse files

Dear deepset,
We are a group of researchers investigating the usefulness of sharing AIBOMs (Artificial Intelligence Bill of Materials) to document AI models and to improve transparency in AI model supply chains. AIBOMs are machine-readable, structured inventories of components鈥攕uch as datasets and models鈥攗sed in the development of AI-powered systems.

We would like to emphasize that we have no financial or competing interests related to AIBOMs. Our sole interest is to advance the collective understanding of AIBOMs within both academia and industry. As part of this effort, we are contributing to randomly selected open and popular models on Hugging Face (like yours) and are happy to offer support to you and the maintainers of your model if needed.

Based on your model card (and some configuration information available in Hugging Face), we generated the AIBOM according to the CyclonDX (v1.6) standard (see https://cyclonedx.org/docs/1.6/json/). This AIBOM is generated as a JSON file by using the following open-source supporting tool: https://github.com/MSR4SBOM/ALOHA (technical details are available in the research paper: https://github.com/MSR4SBOM/ALOHA/blob/main/ALOHA.pdf). This tool is freely available online and can be downloaded and used at your own convenience. We are also happy to assist you directly if you need help generating or reviewing an AIBOM for your model.

The JSON file in this pull request is your AIBOM (see https://github.com/MSR4SBOM/ALOHA/blob/main/documentation.json for details on its structure). Clearly, the submitted AIBOM matches the current model information, yet it can be easily regenerated when the model evolves, using the aforementioned AIBOM generation tool.

We understand that initiatives like ours may raise questions, especially in open communities like Hugging Face. Therefore, we would like to further remark that our interest in AIBOMs is only to enhance the body of knowledge on AIBOMs and to make this easy and low-friction for maintainers of AI models and developers of AI-powered systems.

We open this pull request containing an AIBOM of your AI model, and hope it will be considered. We would also like to hear your opinion on the usefulness (or not) of AIBOM by answering a 3-minute anonymous survey: https://forms.gle/WGffSQD5dLoWttEe7.

Thanks in advance, and regards,
Riccardo D鈥橝vino, Fatima Ahmed, Sabato Nocera, Simone Romano, Giuseppe Scanniello (University of Salerno, Italy),
Massimiliano Di Penta (University of Sannio, Italy),
The MSR4SBOM team

Files changed (1) hide show
  1. deepset_roberta-base-squad2.json +237 -0
deepset_roberta-base-squad2.json ADDED
@@ -0,0 +1,237 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bomFormat": "CycloneDX",
3
+ "specVersion": "1.6",
4
+ "serialNumber": "urn:uuid:cef2615a-e860-42b2-b351-7f8a5f49e535",
5
+ "version": 1,
6
+ "metadata": {
7
+ "timestamp": "2025-07-10T08:45:16.787707+00:00",
8
+ "component": {
9
+ "type": "machine-learning-model",
10
+ "bom-ref": "deepset/roberta-base-squad2-12395755-d71a-5489-a970-16cfa514aa95",
11
+ "name": "deepset/roberta-base-squad2",
12
+ "externalReferences": [
13
+ {
14
+ "url": "https://huggingface.co/deepset/roberta-base-squad2",
15
+ "type": "documentation"
16
+ }
17
+ ],
18
+ "modelCard": {
19
+ "modelParameters": {
20
+ "task": "question-answering",
21
+ "architectureFamily": "roberta",
22
+ "modelArchitecture": "RobertaForQuestionAnswering",
23
+ "datasets": [
24
+ {
25
+ "ref": "squad_v2-9c72005c-340e-5f42-8f7a-ae0c57af7584"
26
+ }
27
+ ]
28
+ },
29
+ "properties": [
30
+ {
31
+ "name": "library_name",
32
+ "value": "transformers"
33
+ },
34
+ {
35
+ "name": "base_model",
36
+ "value": "FacebookAI/roberta-base"
37
+ }
38
+ ],
39
+ "quantitativeAnalysis": {
40
+ "performanceMetrics": [
41
+ {
42
+ "slice": "dataset: squad_v2, split: validation, config: squad_v2",
43
+ "type": "exact_match",
44
+ "value": 79.9309
45
+ },
46
+ {
47
+ "slice": "dataset: squad_v2, split: validation, config: squad_v2",
48
+ "type": "f1",
49
+ "value": 82.9501
50
+ },
51
+ {
52
+ "slice": "dataset: squad_v2, split: validation, config: squad_v2",
53
+ "type": "total",
54
+ "value": 11869
55
+ },
56
+ {
57
+ "slice": "dataset: squad, split: validation, config: plain_text",
58
+ "type": "exact_match",
59
+ "value": 85.289
60
+ },
61
+ {
62
+ "slice": "dataset: squad, split: validation, config: plain_text",
63
+ "type": "f1",
64
+ "value": 91.841
65
+ },
66
+ {
67
+ "slice": "dataset: adversarial_qa, split: validation, config: adversarialQA",
68
+ "type": "exact_match",
69
+ "value": 29.5
70
+ },
71
+ {
72
+ "slice": "dataset: adversarial_qa, split: validation, config: adversarialQA",
73
+ "type": "f1",
74
+ "value": 40.367
75
+ },
76
+ {
77
+ "slice": "dataset: squad_adversarial, split: validation, config: AddOneSent",
78
+ "type": "exact_match",
79
+ "value": 78.567
80
+ },
81
+ {
82
+ "slice": "dataset: squad_adversarial, split: validation, config: AddOneSent",
83
+ "type": "f1",
84
+ "value": 84.469
85
+ },
86
+ {
87
+ "slice": "dataset: squadshifts, split: test, config: amazon",
88
+ "type": "exact_match",
89
+ "value": 69.924
90
+ },
91
+ {
92
+ "slice": "dataset: squadshifts, split: test, config: amazon",
93
+ "type": "f1",
94
+ "value": 83.284
95
+ },
96
+ {
97
+ "slice": "dataset: squadshifts, split: test, config: new_wiki",
98
+ "type": "exact_match",
99
+ "value": 81.204
100
+ },
101
+ {
102
+ "slice": "dataset: squadshifts, split: test, config: new_wiki",
103
+ "type": "f1",
104
+ "value": 90.595
105
+ },
106
+ {
107
+ "slice": "dataset: squadshifts, split: test, config: nyt",
108
+ "type": "exact_match",
109
+ "value": 82.931
110
+ },
111
+ {
112
+ "slice": "dataset: squadshifts, split: test, config: nyt",
113
+ "type": "f1",
114
+ "value": 90.756
115
+ },
116
+ {
117
+ "slice": "dataset: squadshifts, split: test, config: reddit",
118
+ "type": "exact_match",
119
+ "value": 71.55
120
+ },
121
+ {
122
+ "slice": "dataset: squadshifts, split: test, config: reddit",
123
+ "type": "f1",
124
+ "value": 82.939
125
+ }
126
+ ]
127
+ }
128
+ },
129
+ "authors": [
130
+ {
131
+ "name": "deepset"
132
+ }
133
+ ],
134
+ "licenses": [
135
+ {
136
+ "license": {
137
+ "id": "CC-BY-4.0",
138
+ "url": "https://spdx.org/licenses/CC-BY-4.0.html"
139
+ }
140
+ }
141
+ ],
142
+ "description": "**Language model:** roberta-base**Language:** English**Downstream-task:** Extractive QA**Training data:** SQuAD 2.0**Eval data:** SQuAD 2.0**Code:** See [an example extractive QA pipeline built with Haystack](https://haystack.deepset.ai/tutorials/34_extractive_qa_pipeline)**Infrastructure**: 4x Tesla v100",
143
+ "tags": [
144
+ "transformers",
145
+ "pytorch",
146
+ "tf",
147
+ "jax",
148
+ "rust",
149
+ "safetensors",
150
+ "roberta",
151
+ "question-answering",
152
+ "en",
153
+ "dataset:squad_v2",
154
+ "base_model:FacebookAI/roberta-base",
155
+ "base_model:finetune:FacebookAI/roberta-base",
156
+ "license:cc-by-4.0",
157
+ "model-index",
158
+ "endpoints_compatible",
159
+ "region:us"
160
+ ]
161
+ }
162
+ },
163
+ "components": [
164
+ {
165
+ "type": "data",
166
+ "bom-ref": "squad_v2-9c72005c-340e-5f42-8f7a-ae0c57af7584",
167
+ "name": "squad_v2",
168
+ "data": [
169
+ {
170
+ "type": "dataset",
171
+ "bom-ref": "squad_v2-9c72005c-340e-5f42-8f7a-ae0c57af7584",
172
+ "name": "squad_v2",
173
+ "contents": {
174
+ "url": "https://huggingface.co/datasets/squad_v2",
175
+ "properties": [
176
+ {
177
+ "name": "task_categories",
178
+ "value": "question-answering"
179
+ },
180
+ {
181
+ "name": "task_ids",
182
+ "value": "open-domain-qa, extractive-qa"
183
+ },
184
+ {
185
+ "name": "language",
186
+ "value": "en"
187
+ },
188
+ {
189
+ "name": "size_categories",
190
+ "value": "100K<n<1M"
191
+ },
192
+ {
193
+ "name": "annotations_creators",
194
+ "value": "crowdsourced"
195
+ },
196
+ {
197
+ "name": "language_creators",
198
+ "value": "crowdsourced"
199
+ },
200
+ {
201
+ "name": "pretty_name",
202
+ "value": "SQuAD2.0"
203
+ },
204
+ {
205
+ "name": "source_datasets",
206
+ "value": "original"
207
+ },
208
+ {
209
+ "name": "paperswithcode_id",
210
+ "value": "squad"
211
+ },
212
+ {
213
+ "name": "configs",
214
+ "value": "Name of the dataset subset: squad_v2 {\"split\": \"train\", \"path\": \"squad_v2/train-*\"}, {\"split\": \"validation\", \"path\": \"squad_v2/validation-*\"}"
215
+ },
216
+ {
217
+ "name": "license",
218
+ "value": "cc-by-sa-4.0"
219
+ }
220
+ ]
221
+ },
222
+ "governance": {
223
+ "owners": [
224
+ {
225
+ "organization": {
226
+ "name": "rajpurkar",
227
+ "url": "https://huggingface.co/rajpurkar"
228
+ }
229
+ }
230
+ ]
231
+ },
232
+ "description": "\n\t\n\t\t\n\t\tDataset Card for SQuAD 2.0\n\t\n\n\n\t\n\t\t\n\t\tDataset Summary\n\t\n\nStanford Question Answering Dataset (SQuAD) is a reading comprehension dataset, consisting of questions posed by crowdworkers on a set of Wikipedia articles, where the answer to every question is a segment of text, or span, from the corresponding reading passage, or the question might be unanswerable.\nSQuAD 2.0 combines the 100,000 questions in SQuAD1.1 with over 50,000 unanswerable questions written adversarially by crowdworkers\u2026 See the full description on the dataset page: https://huggingface.co/datasets/rajpurkar/squad_v2."
233
+ }
234
+ ]
235
+ }
236
+ ]
237
+ }