jithinpothireddy21 commited on
Commit
3baf7c8
·
verified ·
1 Parent(s): c061a63

EASE-Delta Reader: transformers-native claim-vs-evidence classifier

Browse files
Files changed (7) hide show
  1. LICENSE +428 -0
  2. README.md +174 -0
  3. config.json +94 -0
  4. export_check.json +29 -0
  5. model.safetensors +3 -0
  6. tokenizer.json +0 -0
  7. tokenizer_config.json +21 -0
LICENSE ADDED
@@ -0,0 +1,428 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ Attribution-ShareAlike 4.0 International
2
+
3
+ =======================================================================
4
+
5
+ Creative Commons Corporation ("Creative Commons") is not a law firm and
6
+ does not provide legal services or legal advice. Distribution of
7
+ Creative Commons public licenses does not create a lawyer-client or
8
+ other relationship. Creative Commons makes its licenses and related
9
+ information available on an "as-is" basis. Creative Commons gives no
10
+ warranties regarding its licenses, any material licensed under their
11
+ terms and conditions, or any related information. Creative Commons
12
+ disclaims all liability for damages resulting from their use to the
13
+ fullest extent possible.
14
+
15
+ Using Creative Commons Public Licenses
16
+
17
+ Creative Commons public licenses provide a standard set of terms and
18
+ conditions that creators and other rights holders may use to share
19
+ original works of authorship and other material subject to copyright
20
+ and certain other rights specified in the public license below. The
21
+ following considerations are for informational purposes only, are not
22
+ exhaustive, and do not form part of our licenses.
23
+
24
+ Considerations for licensors: Our public licenses are
25
+ intended for use by those authorized to give the public
26
+ permission to use material in ways otherwise restricted by
27
+ copyright and certain other rights. Our licenses are
28
+ irrevocable. Licensors should read and understand the terms
29
+ and conditions of the license they choose before applying it.
30
+ Licensors should also secure all rights necessary before
31
+ applying our licenses so that the public can reuse the
32
+ material as expected. Licensors should clearly mark any
33
+ material not subject to the license. This includes other CC-
34
+ licensed material, or material used under an exception or
35
+ limitation to copyright. More considerations for licensors:
36
+ wiki.creativecommons.org/Considerations_for_licensors
37
+
38
+ Considerations for the public: By using one of our public
39
+ licenses, a licensor grants the public permission to use the
40
+ licensed material under specified terms and conditions. If
41
+ the licensor's permission is not necessary for any reason--for
42
+ example, because of any applicable exception or limitation to
43
+ copyright--then that use is not regulated by the license. Our
44
+ licenses grant only permissions under copyright and certain
45
+ other rights that a licensor has authority to grant. Use of
46
+ the licensed material may still be restricted for other
47
+ reasons, including because others have copyright or other
48
+ rights in the material. A licensor may make special requests,
49
+ such as asking that all changes be marked or described.
50
+ Although not required by our licenses, you are encouraged to
51
+ respect those requests where reasonable. More considerations
52
+ for the public:
53
+ wiki.creativecommons.org/Considerations_for_licensees
54
+
55
+ =======================================================================
56
+
57
+ Creative Commons Attribution-ShareAlike 4.0 International Public
58
+ License
59
+
60
+ By exercising the Licensed Rights (defined below), You accept and agree
61
+ to be bound by the terms and conditions of this Creative Commons
62
+ Attribution-ShareAlike 4.0 International Public License ("Public
63
+ License"). To the extent this Public License may be interpreted as a
64
+ contract, You are granted the Licensed Rights in consideration of Your
65
+ acceptance of these terms and conditions, and the Licensor grants You
66
+ such rights in consideration of benefits the Licensor receives from
67
+ making the Licensed Material available under these terms and
68
+ conditions.
69
+
70
+
71
+ Section 1 -- Definitions.
72
+
73
+ a. Adapted Material means material subject to Copyright and Similar
74
+ Rights that is derived from or based upon the Licensed Material
75
+ and in which the Licensed Material is translated, altered,
76
+ arranged, transformed, or otherwise modified in a manner requiring
77
+ permission under the Copyright and Similar Rights held by the
78
+ Licensor. For purposes of this Public License, where the Licensed
79
+ Material is a musical work, performance, or sound recording,
80
+ Adapted Material is always produced where the Licensed Material is
81
+ synched in timed relation with a moving image.
82
+
83
+ b. Adapter's License means the license You apply to Your Copyright
84
+ and Similar Rights in Your contributions to Adapted Material in
85
+ accordance with the terms and conditions of this Public License.
86
+
87
+ c. BY-SA Compatible License means a license listed at
88
+ creativecommons.org/compatiblelicenses, approved by Creative
89
+ Commons as essentially the equivalent of this Public License.
90
+
91
+ d. Copyright and Similar Rights means copyright and/or similar rights
92
+ closely related to copyright including, without limitation,
93
+ performance, broadcast, sound recording, and Sui Generis Database
94
+ Rights, without regard to how the rights are labeled or
95
+ categorized. For purposes of this Public License, the rights
96
+ specified in Section 2(b)(1)-(2) are not Copyright and Similar
97
+ Rights.
98
+
99
+ e. Effective Technological Measures means those measures that, in the
100
+ absence of proper authority, may not be circumvented under laws
101
+ fulfilling obligations under Article 11 of the WIPO Copyright
102
+ Treaty adopted on December 20, 1996, and/or similar international
103
+ agreements.
104
+
105
+ f. Exceptions and Limitations means fair use, fair dealing, and/or
106
+ any other exception or limitation to Copyright and Similar Rights
107
+ that applies to Your use of the Licensed Material.
108
+
109
+ g. License Elements means the license attributes listed in the name
110
+ of a Creative Commons Public License. The License Elements of this
111
+ Public License are Attribution and ShareAlike.
112
+
113
+ h. Licensed Material means the artistic or literary work, database,
114
+ or other material to which the Licensor applied this Public
115
+ License.
116
+
117
+ i. Licensed Rights means the rights granted to You subject to the
118
+ terms and conditions of this Public License, which are limited to
119
+ all Copyright and Similar Rights that apply to Your use of the
120
+ Licensed Material and that the Licensor has authority to license.
121
+
122
+ j. Licensor means the individual(s) or entity(ies) granting rights
123
+ under this Public License.
124
+
125
+ k. Share means to provide material to the public by any means or
126
+ process that requires permission under the Licensed Rights, such
127
+ as reproduction, public display, public performance, distribution,
128
+ dissemination, communication, or importation, and to make material
129
+ available to the public including in ways that members of the
130
+ public may access the material from a place and at a time
131
+ individually chosen by them.
132
+
133
+ l. Sui Generis Database Rights means rights other than copyright
134
+ resulting from Directive 96/9/EC of the European Parliament and of
135
+ the Council of 11 March 1996 on the legal protection of databases,
136
+ as amended and/or succeeded, as well as other essentially
137
+ equivalent rights anywhere in the world.
138
+
139
+ m. You means the individual or entity exercising the Licensed Rights
140
+ under this Public License. Your has a corresponding meaning.
141
+
142
+
143
+ Section 2 -- Scope.
144
+
145
+ a. License grant.
146
+
147
+ 1. Subject to the terms and conditions of this Public License,
148
+ the Licensor hereby grants You a worldwide, royalty-free,
149
+ non-sublicensable, non-exclusive, irrevocable license to
150
+ exercise the Licensed Rights in the Licensed Material to:
151
+
152
+ a. reproduce and Share the Licensed Material, in whole or
153
+ in part; and
154
+
155
+ b. produce, reproduce, and Share Adapted Material.
156
+
157
+ 2. Exceptions and Limitations. For the avoidance of doubt, where
158
+ Exceptions and Limitations apply to Your use, this Public
159
+ License does not apply, and You do not need to comply with
160
+ its terms and conditions.
161
+
162
+ 3. Term. The term of this Public License is specified in Section
163
+ 6(a).
164
+
165
+ 4. Media and formats; technical modifications allowed. The
166
+ Licensor authorizes You to exercise the Licensed Rights in
167
+ all media and formats whether now known or hereafter created,
168
+ and to make technical modifications necessary to do so. The
169
+ Licensor waives and/or agrees not to assert any right or
170
+ authority to forbid You from making technical modifications
171
+ necessary to exercise the Licensed Rights, including
172
+ technical modifications necessary to circumvent Effective
173
+ Technological Measures. For purposes of this Public License,
174
+ simply making modifications authorized by this Section 2(a)
175
+ (4) never produces Adapted Material.
176
+
177
+ 5. Downstream recipients.
178
+
179
+ a. Offer from the Licensor -- Licensed Material. Every
180
+ recipient of the Licensed Material automatically
181
+ receives an offer from the Licensor to exercise the
182
+ Licensed Rights under the terms and conditions of this
183
+ Public License.
184
+
185
+ b. Additional offer from the Licensor -- Adapted Material.
186
+ Every recipient of Adapted Material from You
187
+ automatically receives an offer from the Licensor to
188
+ exercise the Licensed Rights in the Adapted Material
189
+ under the conditions of the Adapter's License You apply.
190
+
191
+ c. No downstream restrictions. You may not offer or impose
192
+ any additional or different terms or conditions on, or
193
+ apply any Effective Technological Measures to, the
194
+ Licensed Material if doing so restricts exercise of the
195
+ Licensed Rights by any recipient of the Licensed
196
+ Material.
197
+
198
+ 6. No endorsement. Nothing in this Public License constitutes or
199
+ may be construed as permission to assert or imply that You
200
+ are, or that Your use of the Licensed Material is, connected
201
+ with, or sponsored, endorsed, or granted official status by,
202
+ the Licensor or others designated to receive attribution as
203
+ provided in Section 3(a)(1)(A)(i).
204
+
205
+ b. Other rights.
206
+
207
+ 1. Moral rights, such as the right of integrity, are not
208
+ licensed under this Public License, nor are publicity,
209
+ privacy, and/or other similar personality rights; however, to
210
+ the extent possible, the Licensor waives and/or agrees not to
211
+ assert any such rights held by the Licensor to the limited
212
+ extent necessary to allow You to exercise the Licensed
213
+ Rights, but not otherwise.
214
+
215
+ 2. Patent and trademark rights are not licensed under this
216
+ Public License.
217
+
218
+ 3. To the extent possible, the Licensor waives any right to
219
+ collect royalties from You for the exercise of the Licensed
220
+ Rights, whether directly or through a collecting society
221
+ under any voluntary or waivable statutory or compulsory
222
+ licensing scheme. In all other cases the Licensor expressly
223
+ reserves any right to collect such royalties.
224
+
225
+
226
+ Section 3 -- License Conditions.
227
+
228
+ Your exercise of the Licensed Rights is expressly made subject to the
229
+ following conditions.
230
+
231
+ a. Attribution.
232
+
233
+ 1. If You Share the Licensed Material (including in modified
234
+ form), You must:
235
+
236
+ a. retain the following if it is supplied by the Licensor
237
+ with the Licensed Material:
238
+
239
+ i. identification of the creator(s) of the Licensed
240
+ Material and any others designated to receive
241
+ attribution, in any reasonable manner requested by
242
+ the Licensor (including by pseudonym if
243
+ designated);
244
+
245
+ ii. a copyright notice;
246
+
247
+ iii. a notice that refers to this Public License;
248
+
249
+ iv. a notice that refers to the disclaimer of
250
+ warranties;
251
+
252
+ v. a URI or hyperlink to the Licensed Material to the
253
+ extent reasonably practicable;
254
+
255
+ b. indicate if You modified the Licensed Material and
256
+ retain an indication of any previous modifications; and
257
+
258
+ c. indicate the Licensed Material is licensed under this
259
+ Public License, and include the text of, or the URI or
260
+ hyperlink to, this Public License.
261
+
262
+ 2. You may satisfy the conditions in Section 3(a)(1) in any
263
+ reasonable manner based on the medium, means, and context in
264
+ which You Share the Licensed Material. For example, it may be
265
+ reasonable to satisfy the conditions by providing a URI or
266
+ hyperlink to a resource that includes the required
267
+ information.
268
+
269
+ 3. If requested by the Licensor, You must remove any of the
270
+ information required by Section 3(a)(1)(A) to the extent
271
+ reasonably practicable.
272
+
273
+ b. ShareAlike.
274
+
275
+ In addition to the conditions in Section 3(a), if You Share
276
+ Adapted Material You produce, the following conditions also apply.
277
+
278
+ 1. The Adapter's License You apply must be a Creative Commons
279
+ license with the same License Elements, this version or
280
+ later, or a BY-SA Compatible License.
281
+
282
+ 2. You must include the text of, or the URI or hyperlink to, the
283
+ Adapter's License You apply. You may satisfy this condition
284
+ in any reasonable manner based on the medium, means, and
285
+ context in which You Share Adapted Material.
286
+
287
+ 3. You may not offer or impose any additional or different terms
288
+ or conditions on, or apply any Effective Technological
289
+ Measures to, Adapted Material that restrict exercise of the
290
+ rights granted under the Adapter's License You apply.
291
+
292
+
293
+ Section 4 -- Sui Generis Database Rights.
294
+
295
+ Where the Licensed Rights include Sui Generis Database Rights that
296
+ apply to Your use of the Licensed Material:
297
+
298
+ a. for the avoidance of doubt, Section 2(a)(1) grants You the right
299
+ to extract, reuse, reproduce, and Share all or a substantial
300
+ portion of the contents of the database;
301
+
302
+ b. if You include all or a substantial portion of the database
303
+ contents in a database in which You have Sui Generis Database
304
+ Rights, then the database in which You have Sui Generis Database
305
+ Rights (but not its individual contents) is Adapted Material,
306
+ including for purposes of Section 3(b); and
307
+
308
+ c. You must comply with the conditions in Section 3(a) if You Share
309
+ all or a substantial portion of the contents of the database.
310
+
311
+ For the avoidance of doubt, this Section 4 supplements and does not
312
+ replace Your obligations under this Public License where the Licensed
313
+ Rights include other Copyright and Similar Rights.
314
+
315
+
316
+ Section 5 -- Disclaimer of Warranties and Limitation of Liability.
317
+
318
+ a. UNLESS OTHERWISE SEPARATELY UNDERTAKEN BY THE LICENSOR, TO THE
319
+ EXTENT POSSIBLE, THE LICENSOR OFFERS THE LICENSED MATERIAL AS-IS
320
+ AND AS-AVAILABLE, AND MAKES NO REPRESENTATIONS OR WARRANTIES OF
321
+ ANY KIND CONCERNING THE LICENSED MATERIAL, WHETHER EXPRESS,
322
+ IMPLIED, STATUTORY, OR OTHER. THIS INCLUDES, WITHOUT LIMITATION,
323
+ WARRANTIES OF TITLE, MERCHANTABILITY, FITNESS FOR A PARTICULAR
324
+ PURPOSE, NON-INFRINGEMENT, ABSENCE OF LATENT OR OTHER DEFECTS,
325
+ ACCURACY, OR THE PRESENCE OR ABSENCE OF ERRORS, WHETHER OR NOT
326
+ KNOWN OR DISCOVERABLE. WHERE DISCLAIMERS OF WARRANTIES ARE NOT
327
+ ALLOWED IN FULL OR IN PART, THIS DISCLAIMER MAY NOT APPLY TO YOU.
328
+
329
+ b. TO THE EXTENT POSSIBLE, IN NO EVENT WILL THE LICENSOR BE LIABLE
330
+ TO YOU ON ANY LEGAL THEORY (INCLUDING, WITHOUT LIMITATION,
331
+ NEGLIGENCE) OR OTHERWISE FOR ANY DIRECT, SPECIAL, INDIRECT,
332
+ INCIDENTAL, CONSEQUENTIAL, PUNITIVE, EXEMPLARY, OR OTHER LOSSES,
333
+ COSTS, EXPENSES, OR DAMAGES ARISING OUT OF THIS PUBLIC LICENSE OR
334
+ USE OF THE LICENSED MATERIAL, EVEN IF THE LICENSOR HAS BEEN
335
+ ADVISED OF THE POSSIBILITY OF SUCH LOSSES, COSTS, EXPENSES, OR
336
+ DAMAGES. WHERE A LIMITATION OF LIABILITY IS NOT ALLOWED IN FULL OR
337
+ IN PART, THIS LIMITATION MAY NOT APPLY TO YOU.
338
+
339
+ c. The disclaimer of warranties and limitation of liability provided
340
+ above shall be interpreted in a manner that, to the extent
341
+ possible, most closely approximates an absolute disclaimer and
342
+ waiver of all liability.
343
+
344
+
345
+ Section 6 -- Term and Termination.
346
+
347
+ a. This Public License applies for the term of the Copyright and
348
+ Similar Rights licensed here. However, if You fail to comply with
349
+ this Public License, then Your rights under this Public License
350
+ terminate automatically.
351
+
352
+ b. Where Your right to use the Licensed Material has terminated under
353
+ Section 6(a), it reinstates:
354
+
355
+ 1. automatically as of the date the violation is cured, provided
356
+ it is cured within 30 days of Your discovery of the
357
+ violation; or
358
+
359
+ 2. upon express reinstatement by the Licensor.
360
+
361
+ For the avoidance of doubt, this Section 6(b) does not affect any
362
+ right the Licensor may have to seek remedies for Your violations
363
+ of this Public License.
364
+
365
+ c. For the avoidance of doubt, the Licensor may also offer the
366
+ Licensed Material under separate terms or conditions or stop
367
+ distributing the Licensed Material at any time; however, doing so
368
+ will not terminate this Public License.
369
+
370
+ d. Sections 1, 5, 6, 7, and 8 survive termination of this Public
371
+ License.
372
+
373
+
374
+ Section 7 -- Other Terms and Conditions.
375
+
376
+ a. The Licensor shall not be bound by any additional or different
377
+ terms or conditions communicated by You unless expressly agreed.
378
+
379
+ b. Any arrangements, understandings, or agreements regarding the
380
+ Licensed Material not stated herein are separate from and
381
+ independent of the terms and conditions of this Public License.
382
+
383
+
384
+ Section 8 -- Interpretation.
385
+
386
+ a. For the avoidance of doubt, this Public License does not, and
387
+ shall not be interpreted to, reduce, limit, restrict, or impose
388
+ conditions on any use of the Licensed Material that could lawfully
389
+ be made without permission under this Public License.
390
+
391
+ b. To the extent possible, if any provision of this Public License is
392
+ deemed unenforceable, it shall be automatically reformed to the
393
+ minimum extent necessary to make it enforceable. If the provision
394
+ cannot be reformed, it shall be severed from this Public License
395
+ without affecting the enforceability of the remaining terms and
396
+ conditions.
397
+
398
+ c. No term or condition of this Public License will be waived and no
399
+ failure to comply consented to unless expressly agreed to by the
400
+ Licensor.
401
+
402
+ d. Nothing in this Public License constitutes or may be interpreted
403
+ as a limitation upon, or waiver of, any privileges and immunities
404
+ that apply to the Licensor or You, including from the legal
405
+ processes of any jurisdiction or authority.
406
+
407
+
408
+ =======================================================================
409
+
410
+ Creative Commons is not a party to its public
411
+ licenses. Notwithstanding, Creative Commons may elect to apply one of
412
+ its public licenses to material it publishes and in those instances
413
+ will be considered the “Licensor.” The text of the Creative Commons
414
+ public licenses is dedicated to the public domain under the CC0 Public
415
+ Domain Dedication. Except for the limited purpose of indicating that
416
+ material is shared under a Creative Commons public license or as
417
+ otherwise permitted by the Creative Commons policies published at
418
+ creativecommons.org/policies, Creative Commons does not authorize the
419
+ use of the trademark "Creative Commons" or any other trademark or logo
420
+ of Creative Commons without its prior written consent including,
421
+ without limitation, in connection with any unauthorized modifications
422
+ to any of its public licenses or any other arrangements,
423
+ understandings, or agreements concerning use of licensed material. For
424
+ the avoidance of doubt, this paragraph does not form part of the
425
+ public licenses.
426
+
427
+ Creative Commons may be contacted at creativecommons.org.
428
+
README.md ADDED
@@ -0,0 +1,174 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-sa-4.0
3
+ language:
4
+ - en
5
+ library_name: transformers
6
+ pipeline_tag: text-classification
7
+ base_model: answerdotai/ModernBERT-large
8
+ datasets:
9
+ - tals/vitaminc
10
+ - nyu-mll/multi_nli
11
+ - alisawuffles/WANLI
12
+ tags:
13
+ - natural-language-inference
14
+ - nli
15
+ - fact-verification
16
+ - fact-checking
17
+ - cross-encoder
18
+ - modernbert
19
+ - rag
20
+ - evidence
21
+ metrics:
22
+ - accuracy
23
+ model-index:
24
+ - name: ease-delta-reader
25
+ results:
26
+ - task:
27
+ type: text-classification
28
+ name: Fact verification
29
+ dataset:
30
+ name: VitaminC
31
+ type: tals/vitaminc
32
+ split: test
33
+ metrics:
34
+ - type: accuracy
35
+ value: 91.53
36
+ name: Accuracy
37
+ - task:
38
+ type: text-classification
39
+ name: Natural language inference
40
+ dataset:
41
+ name: MultiNLI (mismatched)
42
+ type: nyu-mll/multi_nli
43
+ split: validation_mismatched
44
+ metrics:
45
+ - type: accuracy
46
+ value: 90.21
47
+ name: Accuracy
48
+ - task:
49
+ type: text-classification
50
+ name: Natural language inference
51
+ dataset:
52
+ name: WANLI
53
+ type: alisawuffles/WANLI
54
+ split: test
55
+ metrics:
56
+ - type: accuracy
57
+ value: 77.42
58
+ name: Accuracy
59
+ - task:
60
+ type: text-classification
61
+ name: Natural language inference
62
+ dataset:
63
+ name: SNLI
64
+ type: stanfordnlp/snli
65
+ split: test
66
+ metrics:
67
+ - type: accuracy
68
+ value: 84.39
69
+ name: Accuracy
70
+ ---
71
+
72
+ # EASE-Delta Reader (large)
73
+
74
+ **Does this passage support the claim, refute it, or not settle it at all?**
75
+
76
+ A 395M-parameter cross-encoder (`answerdotai/ModernBERT-large` fine-tuned) that reads one claim against one passage and answers `SUPPORTS`, `REFUTES` or `NOT_ENOUGH_INFO`. It was trained to say `NOT_ENOUGH_INFO` when a passage is about something else, which is where general NLI models most often guess: on passages taken from unrelated documents it gives a decisive answer 0.58% of the time; a widely used public NLI model does so 45.07% of the time.
77
+
78
+ It is the reader inside [EASE-Delta](https://github.com/jithinsaireddy/ease-delta), a system that keeps a task's decisions current as its messages and documents change. Try both in the [demo](https://huggingface.co/spaces/jithinpothireddy21/ease-delta-demo).
79
+
80
+ ## Use it
81
+
82
+ Claim first, evidence second.
83
+
84
+ ```python
85
+ from transformers import pipeline
86
+
87
+ reader = pipeline("text-classification", model="jithinpothireddy21/ease-delta-reader", top_k=None)
88
+ claim = "The client has approved the final design."
89
+ passages = [
90
+ "Email from the client: we approve the final design, please go ahead.",
91
+ "Email from the client: we cannot approve the design yet; the colours are wrong.",
92
+ "The office is closed on the first Monday of the month for staff training.",
93
+ ]
94
+ for passage in passages:
95
+ print(reader({"text": claim, "text_pair": passage}, truncation=True)[0])
96
+
97
+ # {'label': 'SUPPORTS', 'score': 0.935}
98
+ # {'label': 'REFUTES', 'score': 0.991}
99
+ # {'label': 'NOT_ENOUGH_INFO', 'score': 0.999}
100
+ ```
101
+
102
+ With sentence-transformers, continuing from above:
103
+
104
+ ```python
105
+ from sentence_transformers import CrossEncoder
106
+
107
+ model = CrossEncoder("jithinpothireddy21/ease-delta-reader")
108
+ probs = model.predict([(claim, passage) for passage in passages], apply_softmax=True)
109
+ # one row per pair; columns SUPPORTS, REFUTES, NOT_ENOUGH_INFO
110
+ ```
111
+
112
+ Inputs were at most 256 tokens in training (`truncation=True` cuts there). Split long documents into passages and read each one; say who is speaking in the passage itself ("Email from the client: ..."), because the reader sees only the text.
113
+
114
+ Probabilities are best calibrated after dividing the logits by **1.203** (fitted on development data) before the softmax.
115
+
116
+ ## Results
117
+
118
+ Test sets, accuracy. One evaluation harness for all three models; brackets and paired tests are in the [results report](https://github.com/jithinsaireddy/ease-delta/blob/main/docs/RESULTS.md).
119
+
120
+ | | **This model** (large) | base reader | `tasksource/ModernBERT-base-nli` |
121
+ |---|---:|---:|---:|
122
+ | Passages from unrelated documents read as decisive (lower is better) | 0.58% | 0.82% | 45.07% |
123
+ | VitaminC: claims against Wikipedia revisions | 91.53% | 90.20% | 68.30% |
124
+ | MultiNLI, mismatched genres | 90.21% | 88.56% | 89.84% |
125
+ | WANLI | 77.42% | 74.74% | 66.34% |
126
+ | SNLI (not trained on) | 84.39% | 80.33% | 89.15% |
127
+ | ANLI round 1 (not trained on) | 53.70% | 46.40% | 63.40% |
128
+ | ANLI round 2 (not trained on) | 38.10% | 33.30% | 48.20% |
129
+ | ANLI round 3 (not trained on) | 36.00% | 34.75% | 42.25% |
130
+
131
+ `tasksource/ModernBERT-base-nli` is a base-size multi-task NLI model whose training includes SNLI and ANLI; it is the better choice for adversarial NLI of the ANLI kind. This reader is built for evidence: contrastive Wikipedia revisions (VitaminC), and knowing when a passage does not bear on the claim.
132
+
133
+ On VitaminC, real revisions: 89.90%; synthetic: 94.25%.
134
+
135
+ ## Labels
136
+
137
+ | Label | Meaning | NLI equivalent |
138
+ |---|---|---|
139
+ | `SUPPORTS` | the passage establishes the claim | entailment |
140
+ | `REFUTES` | the passage establishes that the claim is false | contradiction |
141
+ | `NOT_ENOUGH_INFO` | the passage does not settle the claim, including when it is about something else | neutral |
142
+
143
+ ## Training
144
+
145
+ 900,000 examples streamed from the Hugging Face Hub: VitaminC (cc-by-sa-3.0), MultiNLI and WANLI (cc-by-4.0), plus 20% synthetic unrelated pairs (the claim of one example with the evidence of another from a different topic, labelled `NOT_ENOUGH_INFO`). SNLI was left out because its annotators labelled unrelated content as contradiction; ANLI because its licence is non-commercial. One run, on one Apple M4 Max. The evaluation plan was written and hashed before training: [`docs/PREREGISTRATION_LARGE.md`](https://github.com/jithinsaireddy/ease-delta/blob/main/docs/PREREGISTRATION_LARGE.md).
146
+
147
+ ## Limitations
148
+
149
+ - English only.
150
+ - Passages on the claim's own subject that do not settle it are still read as decisive about one time in five (VitaminC test, not-enough-info items).
151
+ - Conflicts that follow only from a consequence ("broke a leg" against "cycles to work") are mostly missed.
152
+ - Numbers and dates are read as text; compare them with code when they matter.
153
+ - Not a judge of truth: it reports what the passage says, not whether the passage is right.
154
+
155
+ ## How this checkpoint was made
156
+
157
+ Converted exactly from the EASE-Delta edge model (`ease/export.py` in the code repository explains why the conversion is exact). On 1,200 development pairs the largest logit difference between the two was 9.5e-07 and the predicted label never differed (0 disagreements). The full system, with its calibrated rules and regression suites, is [`jithinpothireddy21/ease-delta`](https://huggingface.co/jithinpothireddy21/ease-delta).
158
+
159
+ Other size: [`jithinpothireddy21/ease-delta-reader-base`](https://huggingface.co/jithinpothireddy21/ease-delta-reader-base).
160
+
161
+ ## Licence
162
+
163
+ CC BY-SA 4.0 for the weights, because VitaminC is share-alike (Creative Commons' guidance on AI training describes this as the cautious course). The code is Apache-2.0. Credit for the training data: VitaminC (Schuster, Fisch and Barzilay, NAACL 2021), MultiNLI (Williams, Nangia and Bowman, NAACL 2018), WANLI (Liu, Swayamdipta, Smith and Choi, EMNLP 2022). Backbone: ModernBERT (Warner et al., 2024).
164
+
165
+ ## Citation
166
+
167
+ ```bibtex
168
+ @software{pothireddy2026easedelta,
169
+ author = {Pothireddy, Jithin},
170
+ title = {EASE-Delta: revision-aware decision computation},
171
+ year = {2026},
172
+ url = {https://github.com/jithinsaireddy/ease-delta}
173
+ }
174
+ ```
config.json ADDED
@@ -0,0 +1,94 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "ModernBertForSequenceClassification"
4
+ ],
5
+ "attention_bias": false,
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 50281,
8
+ "classifier_activation": "gelu",
9
+ "classifier_bias": true,
10
+ "classifier_dropout": 0.0,
11
+ "classifier_pooling": "cls",
12
+ "cls_token_id": 50281,
13
+ "decoder_bias": true,
14
+ "deterministic_flash_attn": false,
15
+ "dtype": "float32",
16
+ "ease_max_length": 256,
17
+ "embedding_dropout": 0.0,
18
+ "eos_token_id": 50282,
19
+ "global_attn_every_n_layers": 3,
20
+ "gradient_checkpointing": false,
21
+ "hidden_activation": "gelu",
22
+ "hidden_size": 1024,
23
+ "id2label": {
24
+ "0": "SUPPORTS",
25
+ "1": "REFUTES",
26
+ "2": "NOT_ENOUGH_INFO"
27
+ },
28
+ "initializer_cutoff_factor": 2.0,
29
+ "initializer_range": 0.02,
30
+ "intermediate_size": 2624,
31
+ "label2id": {
32
+ "NOT_ENOUGH_INFO": 2,
33
+ "REFUTES": 1,
34
+ "SUPPORTS": 0
35
+ },
36
+ "layer_norm_eps": 1e-05,
37
+ "layer_types": [
38
+ "full_attention",
39
+ "sliding_attention",
40
+ "sliding_attention",
41
+ "full_attention",
42
+ "sliding_attention",
43
+ "sliding_attention",
44
+ "full_attention",
45
+ "sliding_attention",
46
+ "sliding_attention",
47
+ "full_attention",
48
+ "sliding_attention",
49
+ "sliding_attention",
50
+ "full_attention",
51
+ "sliding_attention",
52
+ "sliding_attention",
53
+ "full_attention",
54
+ "sliding_attention",
55
+ "sliding_attention",
56
+ "full_attention",
57
+ "sliding_attention",
58
+ "sliding_attention",
59
+ "full_attention",
60
+ "sliding_attention",
61
+ "sliding_attention",
62
+ "full_attention",
63
+ "sliding_attention",
64
+ "sliding_attention",
65
+ "full_attention"
66
+ ],
67
+ "local_attention": 128,
68
+ "max_position_embeddings": 8192,
69
+ "mlp_bias": false,
70
+ "mlp_dropout": 0.0,
71
+ "model_type": "modernbert",
72
+ "norm_bias": false,
73
+ "norm_eps": 1e-05,
74
+ "num_attention_heads": 16,
75
+ "num_hidden_layers": 28,
76
+ "pad_token_id": 50283,
77
+ "position_embedding_type": "absolute",
78
+ "rope_parameters": {
79
+ "full_attention": {
80
+ "rope_theta": 160000.0,
81
+ "rope_type": "default"
82
+ },
83
+ "sliding_attention": {
84
+ "rope_theta": 10000.0,
85
+ "rope_type": "default"
86
+ }
87
+ },
88
+ "sep_token_id": 50282,
89
+ "sparse_pred_ignore_index": -100,
90
+ "sparse_prediction": false,
91
+ "tie_word_embeddings": true,
92
+ "transformers_version": "5.17.0",
93
+ "vocab_size": 50368
94
+ }
export_check.json ADDED
@@ -0,0 +1,29 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "argmax_disagreements": 0,
3
+ "device": "mps",
4
+ "max_abs_logit_difference": 9.5367431640625e-07,
5
+ "pairs": 1200,
6
+ "seconds": 21.839622974395752,
7
+ "sets": {
8
+ "mnli.dev": {
9
+ "argmax_disagreements": 0,
10
+ "max_abs_logit_difference": 9.5367431640625e-07,
11
+ "pairs": 300
12
+ },
13
+ "snli.dev": {
14
+ "argmax_disagreements": 0,
15
+ "max_abs_logit_difference": 9.5367431640625e-07,
16
+ "pairs": 300
17
+ },
18
+ "unrelated.dev": {
19
+ "argmax_disagreements": 0,
20
+ "max_abs_logit_difference": 9.5367431640625e-07,
21
+ "pairs": 300
22
+ },
23
+ "vitaminc.dev": {
24
+ "argmax_disagreements": 0,
25
+ "max_abs_logit_difference": 9.5367431640625e-07,
26
+ "pairs": 300
27
+ }
28
+ }
29
+ }
model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f8e51d8da318ebc3d2768cf267844480664d34bb8752f7e3ae4a93eb2db6e19
3
+ size 1583359908
tokenizer.json ADDED
The diff for this file is too large to render. See raw diff
 
tokenizer_config.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "backend": "tokenizers",
3
+ "clean_up_tokenization_spaces": true,
4
+ "cls_token": "[CLS]",
5
+ "is_local": true,
6
+ "local_files_only": false,
7
+ "mask_token": "[MASK]",
8
+ "max_length": 256,
9
+ "model_input_names": [
10
+ "input_ids",
11
+ "attention_mask"
12
+ ],
13
+ "model_max_length": 256,
14
+ "pad_token": "[PAD]",
15
+ "sep_token": "[SEP]",
16
+ "stride": 0,
17
+ "tokenizer_class": "TokenizersBackend",
18
+ "truncation_side": "right",
19
+ "truncation_strategy": "longest_first",
20
+ "unk_token": "[UNK]"
21
+ }