DarianNLP commited on
Commit
0bca2cc
·
verified ·
1 Parent(s): 2953557

Upload dataset

Browse files
Files changed (2) hide show
  1. README.md +381 -0
  2. data/train-00000-of-00001.parquet +3 -0
README.md ADDED
@@ -0,0 +1,381 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ dataset_info:
3
+ features:
4
+ - name: idx
5
+ dtype: int64
6
+ - name: prompt
7
+ dtype: string
8
+ - name: label
9
+ dtype: string
10
+ - name: source
11
+ dtype: string
12
+ - name: grad_norm
13
+ dtype: float64
14
+ - name: response
15
+ dtype: string
16
+ - name: delta_Y_mean
17
+ dtype: float64
18
+ - name: delta_Y_per_prompt
19
+ list: float64
20
+ - name: delta_Y_frac_increasing
21
+ dtype: float64
22
+ - name: delta_h10_per_prompt_norms
23
+ list: float64
24
+ - name: delta_h10_mean_norm
25
+ dtype: float64
26
+ - name: delta_f_per_prompt
27
+ struct:
28
+ - name: 10000_direct_inquiries_and_tasks_harmless
29
+ list: float64
30
+ - name: 10737_explaining_concepts_or_skills_harmless
31
+ list: float64
32
+ - name: 1083_direct_problem_solving_queries_harmless
33
+ list: float64
34
+ - name: 10878_directed_fictional_content_generation_harmful
35
+ list: float64
36
+ - name: 10940_request_for_negative_group_stereotypes_harmful
37
+ list: float64
38
+ - name: 11007_requests_for_sensitive_details_harmful
39
+ list: float64
40
+ - name: 11223_metaphorical_aggression_for_positive_goals_harmless
41
+ list: float64
42
+ - name: 11236_prompting_group_denigration_harmful
43
+ list: float64
44
+ - name: 11382_factual_and_explanatory_content_requests_harmless
45
+ list: float64
46
+ - name: 11430_harmful_prescriptive_medical_advice_harmful
47
+ list: float64
48
+ - name: 11634_specific_information_content_requests_harmless
49
+ list: float64
50
+ - name: 11761_specific_instructional_prompts_harmful
51
+ list: float64
52
+ - name: 11861_direct_task_information_request_harmless
53
+ list: float64
54
+ - name: 12102_structured_information_request_harmful
55
+ list: float64
56
+ - name: 12232_specific_information_procedure_requests_harmless
57
+ list: float64
58
+ - name: 1227_harmful_framing_or_instruction_harmful
59
+ list: float64
60
+ - name: 1581_potentially_problematic_information_requests_harmful
61
+ list: float64
62
+ - name: 1636_requesting_sensitive_private_data_harmful
63
+ list: float64
64
+ - name: 1728_task_oriented_communication_processing_harmless
65
+ list: float64
66
+ - name: 1742_general_assistant_tasks_harmless
67
+ list: float64
68
+ - name: 1836_benign_informational_analytical_requests_harmless
69
+ list: float64
70
+ - name: 2197_requests_for_exploitation_tactics_harmful
71
+ list: float64
72
+ - name: 2827_language_requiring_scrutiny_harmless
73
+ list: float64
74
+ - name: 286_corporate_contractual_financial_specifics_harmless
75
+ list: float64
76
+ - name: 2945_harmless_procedural_technical_tasks_harmless
77
+ list: float64
78
+ - name: 3183_instructional_text_operations_harmless
79
+ list: float64
80
+ - name: 3248_crafting_deceptive_exploitative_content_harmful
81
+ list: float64
82
+ - name: 3277_knowledge_and_explanation_queries_harmless
83
+ list: float64
84
+ - name: 3368_guidance_for_self_improvement_understanding_harmless
85
+ list: float64
86
+ - name: 3536_concise_information_extraction_harmless
87
+ list: float64
88
+ - name: 3767_explicit_content_boundary_requests_harmless
89
+ list: float64
90
+ - name: 3772_facilitating_misguidance_or_illicit_acts_harmful
91
+ list: float64
92
+ - name: 3868_request_for_public_address_harmless
93
+ list: float64
94
+ - name: 3895_direct_factual_retrieval_task_harmless
95
+ list: float64
96
+ - name: 3915_harmful_instructional_content_harmful
97
+ list: float64
98
+ - name: 4027_enumeration_and_list_prompts_harmless
99
+ list: float64
100
+ - name: 4112_serious_factual_inquiry_harmless
101
+ list: float64
102
+ - name: 4202_communication_&_information_security_harmless
103
+ list: float64
104
+ - name: 4205_personal_difficulties_and_coping_harmless
105
+ list: float64
106
+ - name: 4211_generating_group_based_discrimination_harmful
107
+ list: float64
108
+ - name: 4319_specific,_direct_inquiry_instruction_harmless
109
+ list: float64
110
+ - name: 4396_publicly_disclosed_personal_health_harmful
111
+ list: float64
112
+ - name: 4528_organizational_risk_communications_harmful
113
+ list: float64
114
+ - name: 4590_soliciting_discriminatory_narratives_harmless
115
+ list: float64
116
+ - name: 4644_specific_information_retrieval_harmless
117
+ list: float64
118
+ - name: 4991_conspiracy_theory_generation_promotion_harmful
119
+ list: float64
120
+ - name: 5449_request_for_email_subject_lines_harmful
121
+ list: float64
122
+ - name: 5505_email_subject_line_task_harmless
123
+ list: float64
124
+ - name: 5509_simple_linguistic_task_prompts_harmless
125
+ list: float64
126
+ - name: 573_specific_information_or_action_request_harmful
127
+ list: float64
128
+ - name: 6090_information_management_and_analysis_harmless
129
+ list: float64
130
+ - name: 6238_factual_definition_queries_harmless
131
+ list: float64
132
+ - name: 6469_academic_advocacy_content_task_harmless
133
+ list: float64
134
+ - name: 6499_constructive_guidance_seeking_harmless
135
+ list: float64
136
+ - name: 6562_direct_task_information_seeking_harmless
137
+ list: float64
138
+ - name: 6733_content_&_information_integrity_harmful
139
+ list: float64
140
+ - name: 6815_generating_influential_content_harmful
141
+ list: float64
142
+ - name: 7063_directive_explanations_&_creations_harmless
143
+ list: float64
144
+ - name: 7100_requests_for_specific_actionable_data_harmful
145
+ list: float64
146
+ - name: 7261_seeking_safety_and_protection_guidance_harmless
147
+ list: float64
148
+ - name: 7294_propagating_harmful_or_fraudulent_actions_harmful
149
+ list: float64
150
+ - name: 7480_email_subject_line_query_harmless
151
+ list: float64
152
+ - name: 7565_instructional_content_generation_harmless
153
+ list: float64
154
+ - name: 7575_explicit_information_requests_harmless
155
+ list: float64
156
+ - name: 7677_constrained_text_generation_instruction_harmless
157
+ list: float64
158
+ - name: 7897_action_oriented_practical_queries_harmless
159
+ list: float64
160
+ - name: 7940_strategic_deception_or_corporate_manipulation_harmful
161
+ list: float64
162
+ - name: 7971_factual_entity_information_queries_harmless
163
+ list: float64
164
+ - name: 8063_noteworthy_entities'_positive_ethical_aspects_harmless
165
+ list: float64
166
+ - name: 808_promoting_group_discrimination_by_traits_harmful
167
+ list: float64
168
+ - name: 8193_detailed_content_generation_requests_harmful
169
+ list: float64
170
+ - name: 823_instructions_for_mathematical_tasks_harmless
171
+ list: float64
172
+ - name: 8240_explicit_structured_information_tasks_harmless
173
+ list: float64
174
+ - name: 8643_neutral_framed_content_tasks_harmless
175
+ list: float64
176
+ - name: 9081_managing_sensitive_critical_information_harmful
177
+ list: float64
178
+ - name: 910_email_subject_line_task_harmless
179
+ list: float64
180
+ - name: 9267_contextually_harmless_fictional_scenarios_harmless
181
+ list: float64
182
+ - name: 9623_disinformation_for_manipulation_harmful
183
+ list: float64
184
+ - name: 9649_**soliciting_harmful_content_creation**_harmful
185
+ list: float64
186
+ - name: 9994_requests_for_n_item_lists_harmless
187
+ list: float64
188
+ - name: delta_f_per_prompt_norms
189
+ list: float64
190
+ - name: delta_f_mean_norm
191
+ dtype: float64
192
+ - name: refusal_influence
193
+ dtype: float64
194
+ - name: influence_10000_direct_inquiries_and_tasks_harmless
195
+ dtype: float64
196
+ - name: influence_10737_explaining_concepts_or_skills_harmless
197
+ dtype: float64
198
+ - name: influence_1083_direct_problem_solving_queries_harmless
199
+ dtype: float64
200
+ - name: influence_10878_directed_fictional_content_generation_harmful
201
+ dtype: float64
202
+ - name: influence_10940_request_for_negative_group_stereotypes_harmful
203
+ dtype: float64
204
+ - name: influence_11007_requests_for_sensitive_details_harmful
205
+ dtype: float64
206
+ - name: influence_11223_metaphorical_aggression_for_positive_goals_harmless
207
+ dtype: float64
208
+ - name: influence_11236_prompting_group_denigration_harmful
209
+ dtype: float64
210
+ - name: influence_11382_factual_and_explanatory_content_requests_harmless
211
+ dtype: float64
212
+ - name: influence_11430_harmful_prescriptive_medical_advice_harmful
213
+ dtype: float64
214
+ - name: influence_11634_specific_information_content_requests_harmless
215
+ dtype: float64
216
+ - name: influence_11761_specific_instructional_prompts_harmful
217
+ dtype: float64
218
+ - name: influence_11861_direct_task_information_request_harmless
219
+ dtype: float64
220
+ - name: influence_12102_structured_information_request_harmful
221
+ dtype: float64
222
+ - name: influence_12232_specific_information_procedure_requests_harmless
223
+ dtype: float64
224
+ - name: influence_1227_harmful_framing_or_instruction_harmful
225
+ dtype: float64
226
+ - name: influence_1581_potentially_problematic_information_requests_harmful
227
+ dtype: float64
228
+ - name: influence_1636_requesting_sensitive_private_data_harmful
229
+ dtype: float64
230
+ - name: influence_1728_task_oriented_communication_processing_harmless
231
+ dtype: float64
232
+ - name: influence_1742_general_assistant_tasks_harmless
233
+ dtype: float64
234
+ - name: influence_1836_benign_informational_analytical_requests_harmless
235
+ dtype: float64
236
+ - name: influence_2197_requests_for_exploitation_tactics_harmful
237
+ dtype: float64
238
+ - name: influence_2827_language_requiring_scrutiny_harmless
239
+ dtype: float64
240
+ - name: influence_286_corporate_contractual_financial_specifics_harmless
241
+ dtype: float64
242
+ - name: influence_2945_harmless_procedural_technical_tasks_harmless
243
+ dtype: float64
244
+ - name: influence_3183_instructional_text_operations_harmless
245
+ dtype: float64
246
+ - name: influence_3248_crafting_deceptive_exploitative_content_harmful
247
+ dtype: float64
248
+ - name: influence_3277_knowledge_and_explanation_queries_harmless
249
+ dtype: float64
250
+ - name: influence_3368_guidance_for_self_improvement_understanding_harmless
251
+ dtype: float64
252
+ - name: influence_3536_concise_information_extraction_harmless
253
+ dtype: float64
254
+ - name: influence_3767_explicit_content_boundary_requests_harmless
255
+ dtype: float64
256
+ - name: influence_3772_facilitating_misguidance_or_illicit_acts_harmful
257
+ dtype: float64
258
+ - name: influence_3868_request_for_public_address_harmless
259
+ dtype: float64
260
+ - name: influence_3895_direct_factual_retrieval_task_harmless
261
+ dtype: float64
262
+ - name: influence_3915_harmful_instructional_content_harmful
263
+ dtype: float64
264
+ - name: influence_4027_enumeration_and_list_prompts_harmless
265
+ dtype: float64
266
+ - name: influence_4112_serious_factual_inquiry_harmless
267
+ dtype: float64
268
+ - name: influence_4202_communication_&_information_security_harmless
269
+ dtype: float64
270
+ - name: influence_4205_personal_difficulties_and_coping_harmless
271
+ dtype: float64
272
+ - name: influence_4211_generating_group_based_discrimination_harmful
273
+ dtype: float64
274
+ - name: influence_4319_specific,_direct_inquiry_instruction_harmless
275
+ dtype: float64
276
+ - name: influence_4396_publicly_disclosed_personal_health_harmful
277
+ dtype: float64
278
+ - name: influence_4528_organizational_risk_communications_harmful
279
+ dtype: float64
280
+ - name: influence_4590_soliciting_discriminatory_narratives_harmless
281
+ dtype: float64
282
+ - name: influence_4644_specific_information_retrieval_harmless
283
+ dtype: float64
284
+ - name: influence_4991_conspiracy_theory_generation_promotion_harmful
285
+ dtype: float64
286
+ - name: influence_5449_request_for_email_subject_lines_harmful
287
+ dtype: float64
288
+ - name: influence_5505_email_subject_line_task_harmless
289
+ dtype: float64
290
+ - name: influence_5509_simple_linguistic_task_prompts_harmless
291
+ dtype: float64
292
+ - name: influence_573_specific_information_or_action_request_harmful
293
+ dtype: float64
294
+ - name: influence_6090_information_management_and_analysis_harmless
295
+ dtype: float64
296
+ - name: influence_6238_factual_definition_queries_harmless
297
+ dtype: float64
298
+ - name: influence_6469_academic_advocacy_content_task_harmless
299
+ dtype: float64
300
+ - name: influence_6499_constructive_guidance_seeking_harmless
301
+ dtype: float64
302
+ - name: influence_6562_direct_task_information_seeking_harmless
303
+ dtype: float64
304
+ - name: influence_6733_content_&_information_integrity_harmful
305
+ dtype: float64
306
+ - name: influence_6815_generating_influential_content_harmful
307
+ dtype: float64
308
+ - name: influence_7063_directive_explanations_&_creations_harmless
309
+ dtype: float64
310
+ - name: influence_7100_requests_for_specific_actionable_data_harmful
311
+ dtype: float64
312
+ - name: influence_7261_seeking_safety_and_protection_guidance_harmless
313
+ dtype: float64
314
+ - name: influence_7294_propagating_harmful_or_fraudulent_actions_harmful
315
+ dtype: float64
316
+ - name: influence_7480_email_subject_line_query_harmless
317
+ dtype: float64
318
+ - name: influence_7565_instructional_content_generation_harmless
319
+ dtype: float64
320
+ - name: influence_7575_explicit_information_requests_harmless
321
+ dtype: float64
322
+ - name: influence_7677_constrained_text_generation_instruction_harmless
323
+ dtype: float64
324
+ - name: influence_7897_action_oriented_practical_queries_harmless
325
+ dtype: float64
326
+ - name: influence_7940_strategic_deception_or_corporate_manipulation_harmful
327
+ dtype: float64
328
+ - name: influence_7971_factual_entity_information_queries_harmless
329
+ dtype: float64
330
+ - name: influence_8063_noteworthy_entities'_positive_ethical_aspects_harmless
331
+ dtype: float64
332
+ - name: influence_808_promoting_group_discrimination_by_traits_harmful
333
+ dtype: float64
334
+ - name: influence_8193_detailed_content_generation_requests_harmful
335
+ dtype: float64
336
+ - name: influence_823_instructions_for_mathematical_tasks_harmless
337
+ dtype: float64
338
+ - name: influence_8240_explicit_structured_information_tasks_harmless
339
+ dtype: float64
340
+ - name: influence_8643_neutral_framed_content_tasks_harmless
341
+ dtype: float64
342
+ - name: influence_9081_managing_sensitive_critical_information_harmful
343
+ dtype: float64
344
+ - name: influence_910_email_subject_line_task_harmless
345
+ dtype: float64
346
+ - name: influence_9267_contextually_harmless_fictional_scenarios_harmless
347
+ dtype: float64
348
+ - name: influence_9623_disinformation_for_manipulation_harmful
349
+ dtype: float64
350
+ - name: influence_9649_**soliciting_harmful_content_creation**_harmful
351
+ dtype: float64
352
+ - name: influence_9994_requests_for_n_item_lists_harmless
353
+ dtype: float64
354
+ - name: top5_most_influenced
355
+ list:
356
+ - name: feature
357
+ dtype: string
358
+ - name: influence
359
+ dtype: float64
360
+ - name: ridge_weight
361
+ dtype: float64
362
+ - name: top5_most_important
363
+ list:
364
+ - name: feature
365
+ dtype: string
366
+ - name: influence
367
+ dtype: float64
368
+ - name: ridge_weight
369
+ dtype: float64
370
+ splits:
371
+ - name: train
372
+ num_bytes: 149828784
373
+ num_examples: 41
374
+ download_size: 150032532
375
+ dataset_size: 149828784
376
+ configs:
377
+ - config_name: default
378
+ data_files:
379
+ - split: train
380
+ path: data/train-*
381
+ ---
data/train-00000-of-00001.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e2a79f0deb9eabd1ef232843a7b6fef8d550357f1596d59c8c60e5f50c00fe53
3
+ size 150032532