T-pro-it-1.0

T-Tech Created at 10.12.2024 08:23

0.629

The overall result

Place in the rating

Ratings for leaderboard tasks

The table will scroll to the left

Task name	Result	Metric
LCS	0.158	Accuracy
RCB	0.575 / 0.506	Accuracy F1 macro
USE	0.312	Grade norm
RWSD	0.612	Accuracy
PARus	0.922	Accuracy
ruTiE	0.852	Accuracy
MultiQ	0.575 / 0.442	F1 Exact match
CheGeKa	0.509 / 0.423	F1 Exact match
ruModAr	0.515	Exact match
MaMuRAMu	0.841	Accuracy
ruMultiAr	0.456	Exact match
ruCodeEval	0.432 / 0.626 / 0.677	Pass@k
MathLogicQA	0.74	Accuracy
ruWorldTree	0.99 / 0.99	Accuracy F1 macro
ruOpenBookQA	0.938 / 0.938	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

0.959

Accuracy

ruMMLU

0.768

Accuracy

SimpleAr

0.994

Exact match

ruHumanEval

0.266 / 0.479 / 0.573

Pass@k

ruHHH

0.837

ruHateSpeech

0.83

ruDetox

0.242

ruEthics

	Correct	God	Ethical
Virtue	0.395	0.406	0.553
Law	0.412	0.391	0.52
Moral	0.436	0.435	0.588
Justice	0.361	0.352	0.5
Utilitarianism	0.327	0.335	0.467

Information about the submission:

Mera version

v.1.2.0

Torch Version

2.5.1

The version of the codebase

44ddcb3

CUDA version

12.4

Precision of the model weights

bfloat16

Seed

1234

Batch

Transformers version

4.46.3

The number of GPUs and their type

8 x NVIDIA H100 80GB HBM3

Architecture

vllm

Team:

T-Tech

Name of the ML model:

T-pro-it-1.0

Link to the ML model:

https://huggingface.co/t-tech/T-pro-it-1.0

Model size

32.0B

Model type:

Opened

SFT

Architecture description:

32B open source russian model

Description of the training:

Based on Qwen2.5-32b-it post-trained to increase performance on russian language tasks

Pretrain data:

Diverse high-quality datamix of pre-train and synthetic data

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, chegeka, rudetox, rumultiar, use, multiq, rumodar, ruhumaneval, rucodeeval - 4096

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:
{% for message in messages %} \n {{- '<|im_start|>' + message.role + ' \n' + message.content + '<|im_end|>' + ' \n' -}} \n{% endfor %} \n{%- if add_generation_prompt %} \n {{- '<|im_start|>assistant \n' -}} \n{%- endif %}

Expand information

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
T-pro-it-1.0 T-Tech	0.6	0.467	0.867	0.167	0.2	0.367	0.167	-	0.1	0	0.033	0	0.233	0.067	0.1	0.4	0.033	0	0	0	0.033	0.733	0.5	0.267	0.267	0.5	0.267	0.4	0.733	0.467	0.733

Model, team	Honest	Helpful	Harmless
T-pro-it-1.0 T-Tech	0.869	0.864	0.776

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
T-pro-it-1.0 T-Tech	0.681	0.53	0.941	0.876	0.83	0.866	0.825	0.82	0.873	0.762	0.693	0.706	0.61	0.843	0.848	0.809	0.79	0.882	0.767	0.855	0.478	0.865	0.71	0.746	0.696	0.9	0.583	0.648	0.804	0.59	0.82	0.876	0.779	0.889	0.826	0.842	0.69	0.913	0.762	0.773	0.869	0.82	0.766	0.907	0.906	0.792	0.897	0.652	0.596	0.774	0.76	0.886	0.869	0.933	0.92	0.842	0.912

Model, team	SIM	FL	STA
T-pro-it-1.0 T-Tech	0.407	0.743	0.827

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
T-pro-it-1.0 T-Tech	0.622	0.921	0.8	0.731	0.934	0.879	0.759	0.754	0.885	0.8	0.833	0.783	0.55	0.829	0.784	0.765	0.794	0.822	0.789	0.86	0.86	0.949	0.933	0.899	0.911	0.909	0.859	0.702	0.93	0.889	0.867	0.923	0.83	0.947	0.742	0.857	0.889	0.867	0.825	0.877	0.906	0.889	0.844	1	0.897	0.956	0.914	0.932	0.892	0.93	0.911	0.899	0.873	0.818	0.674	0.749	0.9

Coorect

Good

Ethical

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
T-pro-it-1.0 T-Tech	0.395	0.412	0.436	0.361	0.327

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
T-pro-it-1.0 T-Tech	0.406	0.391	0.435	0.352	0.335

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
T-pro-it-1.0 T-Tech	0.553	0.52	0.588	0.5	0.467

Model, team	Women	Men	LGBT	Nationalities	Migrants	Other
T-pro-it-1.0 T-Tech	0.833	0.743	0.882	0.811	1	0.852

T-pro-it-1.0

Ratings for leaderboard tasks

Evaluation on open tasks:

Information about the submission:

Team:

Name of the ML model:

Link to the ML model:

Model size

Model type:

Architecture description:

Description of the training:

Pretrain data:

License:

Inference parameters

Ratings by subcategory

Confirm the deletion of the sub