Add dan, ucar, translation, base64, self-refine attacks

[NickoJo]

added/localized DAN, UCAR and Translation jailbreak attacks

[github-actions]

🧪 Test coverage: 0.00%

Code Coverage Summary

Filename                                          Stmts    Miss  Cover    Missing
----------------------------------------------  -------  ------  -------  ---------
llamator/hello.py                                     2       2  0.00%    4-6
llamator/initial_validation.py                       30      30  0.00%    1-92
llamator/main.py                                     34      34  0.00%    2-104
llamator/ps_logging.py                               15      15  0.00%    1-37
llamator/attack_provider/attack_loader.py             1       1  0.00%    1
llamator/attack_provider/attack_registry.py          25      25  0.00%    1-76
llamator/attack_provider/run_tests.py                91      91  0.00%    1-363
llamator/attack_provider/test_base.py                52      52  0.00%    1-111
llamator/attack_provider/util.py                     27      27  0.00%    1-82
llamator/attack_provider/work_progress_pool.py       80      80  0.00%    1-135
llamator/attacks/aim.py                              40      40  0.00%    1-120
llamator/attacks/base64_injection.py                 64      64  0.00%    1-144
llamator/attacks/complimentary_transition.py         39      39  0.00%    1-109
llamator/attacks/dan.py                              56      56  0.00%    1-138
llamator/attacks/dynamic_test.py                     83      83  0.00%    1-244
llamator/attacks/ethical_compliance.py               40      40  0.00%    1-108
llamator/attacks/harmful_behavior.py                 43      43  0.00%    1-98
llamator/attacks/ru_dan.py                           56      56  0.00%    1-133
llamator/attacks/ru_self_refine.py                   57      57  0.00%    1-141
llamator/attacks/ru_ucar.py                          51      51  0.00%    1-121
llamator/attacks/self_refine.py                      57      57  0.00%    1-141
llamator/attacks/sycophancy.py                       77      77  0.00%    1-255
llamator/attacks/translation.py                      39      39  0.00%    1-98
llamator/attacks/typoglycemia.py                     28      28  0.00%    1-50
llamator/attacks/ucar.py                             55      55  0.00%    1-139
llamator/attacks/utils.py                            10      10  0.00%    1-17
llamator/client/attack_config.py                      5       5  0.00%    1-7
llamator/client/chat_client.py                       33      33  0.00%    1-131
llamator/client/client_config.py                     21      21  0.00%    1-39
llamator/client/langchain_integration.py             79      79  0.00%    1-131
llamator/client/specific_chat_clients.py             63      63  0.00%    1-220
llamator/format_output/logo.py                        8       8  0.00%    1-17
llamator/format_output/results_table.py              23      23  0.00%    1-30
TOTAL                                              1384    1384  0.00%

Diff against main

Filename                                Stmts    Miss  Cover
------------------------------------  -------  ------  --------
llamator/attacks/base64_injection.py      +20     +20  +100.00%
llamator/attacks/dan.py                   +16     +16  +100.00%
llamator/attacks/ru_dan.py                +56     +56  +100.00%
llamator/attacks/ru_self_refine.py        +57     +57  +100.00%
llamator/attacks/ru_ucar.py               +51     +51  +100.00%
llamator/attacks/self_refine.py           +17     +17  +100.00%
llamator/attacks/translation.py            -2      -2  +100.00%
llamator/attacks/ucar.py                  +15     +15  +100.00%
TOTAL                                    +230    +230  +100.00%

Results for commit: 3d93486

Minimum allowed coverage is 60%

♻️ This comment has been updated with latest results

[nizamovtimur]

LGTM

[NickoJo]

• докстринги

[NickoJo]

1. К своим существующим атакам добавил функционал логирования из ветки test_artifacts.
2. Помимо переименования локальных атак вернул англоязычные атаки (dan/ru_dan, ucar/ru_ucar).
3. Добавил атаку ru_self_refined и self_refined.
4. В датасет base64 добавил англоязычные промпты.
5. utils.py пополним английскими refusal_words.
6. Обновил attack_loader.py (в список добавил новые ru* атаки)

[NickoJo]

смерджил актуальную ветку main (после изменений @RomiconEZ) с текущей веткой feature/jailbreak_attacks

[nizamovtimur]

!!!

[RomiconEZ]

Проверено