Watari-7b-v1

Attention Signs Created at 21.02.2025 07:23
0.423
The overall result
291
Place in the rating
Weak tasks:
118
RWSD
282
PARus
390
RCB
286
ruEthics
364
MultiQ
253
ruWorldTree
245
ruOpenBookQA
275
CheGeKa
271
ruMMLU
289
ruHateSpeech
185
ruDetox
268
ruHHH
372
ruTiE
243
ruHumanEval
251
USE
202
MathLogicQA
342
ruMultiAr
328
SimpleAr
546
LCS
354
BPS
322
ruModAr
267
MaMuRAMu
288
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.044 Accuracy
RCB 0.484 / 0.401 Accuracy F1 macro
USE 0.16 Grade norm
RWSD 0.592 Accuracy
PARus 0.774 Accuracy
ruTiE 0.572 Accuracy
MultiQ 0.298 / 0.17 F1 Exact match
CheGeKa 0.11 / 0.084 F1 Exact match
ruModAr 0.441 Exact match
MaMuRAMu 0.649 Accuracy
ruMultiAr 0.219 Exact match
ruCodeEval 0.001 / 0.005 / 0.006 Pass@k
MathLogicQA 0.455 Accuracy
ruWorldTree 0.876 / 0.876 Accuracy F1 macro
ruOpenBookQA 0.788 / 0.787 Accuracy F1 macro

Evaluation on open tasks:

Go to the ratings by subcategory

The table will scroll to the left

Task name Result Metric
BPS 0.859 Accuracy
ruMMLU 0.539 Accuracy
SimpleAr 0.923 Exact match
ruHumanEval 0.016 / 0.055 / 0.079 Pass@k
ruHHH 0.657
ruHateSpeech 0.675
ruDetox 0.213
ruEthics
Correct God Ethical
Virtue 0.232 0.378 0.307
Law 0.243 0.387 0.298
Moral 0.275 0.383 0.34
Justice 0.219 0.328 0.256
Utilitarianism 0.2 0.297 0.302

Information about the submission

Mera version
v.1.2.0
Torch Version
2.5.1
The version of the codebase
30667dc
CUDA version
12.4
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.48.2
The number of GPUs and their type
2 x NVIDIA A100 80GB PCIe
Architecture
hf

Team:

Attention Signs

Name of the ML model:

Watari-7b-v1

Model size

7.6B

Model type:

Opened

SFT

Architecture description:

Инстинктивная версия модели Qwen2.5 (7B), дообученная полным циклом SFT на датасете GrandMaster Pro Max

Description of the training:

Full supervised fine-tuning for 1 epoch

Pretrain data:

Vikhrmodels/GrandMaster-PRO-MAX

License:

Apache 2.0

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=["\n"]; \nchegeka - do_sample=false;until=["\n"]; \nrudetox - do_sample=false;until=["\n"]; \nrumultiar - do_sample=false;until=["\n"]; \nuse - do_sample=false;until=["\n","."]; \nmultiq - do_sample=false;until=["\n"]; \nrumodar - do_sample=false;until=["\n"]; \nruhumaneval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6; \nrucodeeval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6;

The size of the context:
simplear, bps, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, ruhumaneval, rucodeeval, rutie - 32768

System prompt:
Реши задачу по инструкции ниже. Не давай никаких объяснений и пояснений к своему ответу. Не пиши ничего лишнего. Пиши только то, что указано в инструкции. Если по инструкции нужно решить пример, то напиши только числовой ответ без хода решения и пояснений. Если по инструкции нужно вывести букву, цифру или слово, выведи только его. Если по инструкции нужно выбрать один из вариантов ответа и вывести букву или цифру, которая ему соответствует, то выведи только эту букву или цифру, не давай никаких пояснений, не добавляй знаки препинания, только 1 символ в ответе. Если по инструкции нужно дописать код функции на языке Python, пиши сразу код, соблюдая отступы так, будто ты продолжаешь функцию из инструкции, не давай пояснений, не пиши комментарии, используй только аргументы из сигнатуры функции в инструкции, не пробуй считывать данные через функцию input. Не извиняйся, не строй диалог. Выдавай только ответ и ничего больше.

Description of the template:
"{% set loop_messages = messages %}{% for message in loop_messages %}{% set content = '<|im_start|>' + message['role'] + '<|im_sep|>'+ message['content'] | trim + '<|im_end|>' %}{{ content }}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant<|im_sep|>' }}{% endif %}"

Ratings by subcategory

Metric: Grade Norm
Model, team 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 8_0 8_1 8_2 8_3 8_4
Watari-7b-v1
Attention Signs
0.267 0.233 0.633 0.167 0.033 0.133 0.033 - 0 0.033 0 0 0.1 0 0.1 0.183 0.067 0 0 0 0 0.5 0.2 0.033 0 0.225 0.133 0.333 0.5 0.267 0.4
Model, team Honest Helpful Harmless
Watari-7b-v1
Attention Signs
0.557 0.729 0.69
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Management Philosophy Prehistory Human aging Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Biology (college) Physics (college) Human Sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine learning Medical genetics Professional law PR Security studies Chemistry (школьная) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual_physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) European history Government and politics
Watari-7b-v1
Attention Signs
0.541 0.446 0.658 0.731 0.621 0.741 0.699 0.595 0.59 0.601 0.377 0.381 0.33 0.611 0.651 0.621 0.64 0.563 0.311 0.557 0.245 0.737 0.34 0.451 0.42 0.55 0.412 0.556 0.653 0.38 0.67 0.686 0.54 0.778 0.615 0.564 0.43 0.703 0.43 0.473 0.702 0.445 0.621 0.507 0.71 0.528 0.676 0.393 0.358 0.493 0.52 0.684 0.562 0.651 0.75 0.673 0.617
Model, team SIM FL STA
Watari-7b-v1
Attention Signs
0.583 0.719 0.618
Model, team Anatomy Virology Astronomy Marketing Nutrition Sociology Managment Philosophy Pre-History Gerontology Econometrics Formal logic Global facts Jurisprudence Miscellaneous Moral disputes Business ethics Bilology (college) Physics (college) Human sexuality Moral scenarios World religions Abstract algebra Medicine (college) Machine Learning Genetics Professional law PR Security Chemistry (college) Computer security International law Logical fallacies Politics Clinical knowledge Conceptual physics Math (college) Biology (high school) Physics (high school) Chemistry (high school) Geography (high school) Professional medicine Electrical Engineering Elementary mathematics Psychology (high school) Statistics (high school) History (high school) Math (high school) Professional Accounting Professional psychology Computer science (college) World history (high school) Macroeconomics Microeconomics Computer science (high school) Europe History Government and politics
Watari-7b-v1
Attention Signs
0.533 0.634 0.483 0.574 0.724 0.655 0.552 0.667 0.635 0.554 0.692 0.7 0.475 0.667 0.561 0.593 0.645 0.689 0.456 0.719 0.298 0.678 0.689 0.71 0.644 0.682 0.654 0.614 0.877 0.711 0.8 0.795 0.616 0.737 0.515 0.732 0.689 0.689 0.526 0.677 0.702 0.635 0.778 0.644 0.897 0.844 0.81 0.727 0.692 0.719 0.689 0.652 0.785 0.519 0.419 0.55 0.678
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Watari-7b-v1
Attention Signs
0.232 0.243 0.275 0.219 0.2
Model, team Virtue Law Moral Justice Utilitarianism
Watari-7b-v1
Attention Signs
0.378 0.387 0.383 0.328 0.297
Model, team Virtue Law Moral Justice Utilitarianism
Watari-7b-v1
Attention Signs
0.307 0.298 0.34 0.256 0.302
Model, team Women Men LGBT Nationalities Migrants Other
Watari-7b-v1
Attention Signs
0.63 0.6 0.882 0.811 0.714 0.656