GigaChat Max

GIGACHAT Created at 24.10.2024 07:30

0.588

The overall result

Place in the rating

Ratings for leaderboard tasks

The table will scroll to the left

Task name	Result	Metric
LCS	0.192	Accuracy
RCB	0.58 / 0.423	Accuracy F1 macro
USE	0.393	Grade norm
RWSD	0.665	Accuracy
PARus	0.928	Accuracy
ruTiE	0.715	Accuracy
MultiQ	0.486 / 0.322	F1 Exact match
CheGeKa	0.469 / 0.397	F1 Exact match
ruModAr	0.938	Exact match
MaMuRAMu	0.824	Accuracy
ruMultiAr	0.362	Exact match
ruCodeEval	0.077 / 0.093 / 0.098	Pass@k
MathLogicQA	0.575	Accuracy
ruWorldTree	0.975 / 0.975	Accuracy F1 macro
ruOpenBookQA	0.918 / 0.737	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

0.977

Accuracy

ruMMLU

0.718

Accuracy

SimpleAr

0.989

Exact match

ruHumanEval

0.184 / 0.195 / 0.201

Pass@k

ruHHH

0.775

ruHateSpeech

0.611

ruDetox

0.199

ruEthics

	Correct	God	Ethical
Virtue	0.394	0.369	0.42
Law	0.379	0.358	0.422
Moral	0.422	0.386	0.451
Justice	0.347	0.328	0.377
Utilitarianism	0.336	0.332	0.386

Information about the submission:

Mera version

v.1.2.0

Torch Version

2.4.0

The version of the codebase

44ddcb3

CUDA version

12.1

Precision of the model weights

Seed

1234

Butch

Transformers version

4.43.2

The number of GPUs and their type

5 x NVIDIA H100 80GB HBM3

Architecture

gigachat_llms

Team:

GIGACHAT

Name of the ML model:

GigaChat Max

Link to the ML model:

https://developers.sber.ru/docs/ru/gigachat/models

Model type:

Closed

API

Architecture description:

GigaChat MAX (version 1.0.0.0) is the largest in GigaChat Model family (LLM). The model will be available for B2C users https://giga.chat/ and B2B users via API soon https://developers.sber.ru/docs/ru/gigachat/api/tariffs.

Description of the training:

—

Pretrain data:

—

License:

Proprietary model by Sber

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"]; \nrucodeeval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];

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:
API Default

Expand information

Ratings by subcategory

Metric: Accuracy

Model, team	Honest	Helpful	Harmless
GigaChat Max GIGACHAT	0.77	0.729	0.828

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
GigaChat Max GIGACHAT	0.659	0.524	0.908	0.855	0.85	0.821	0.806	0.775	0.846	0.735	0.605	0.556	0.47	0.741	0.851	0.769	0.76	0.861	0.567	0.786	0.642	0.865	0.42	0.711	0.67	0.87	0.505	0.648	0.776	0.52	0.77	0.851	0.748	0.879	0.751	0.774	0.47	0.897	0.563	0.665	0.874	0.794	0.724	0.594	0.884	0.648	0.814	0.437	0.528	0.739	0.67	0.785	0.779	0.84	0.83	0.818	0.886

Model, team	SIM	FL	STA
GigaChat Max GIGACHAT	0.327	0.749	0.765

Coorect

Good

Ethical

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Max GIGACHAT	0.394	0.379	0.422	0.347	0.336

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Max GIGACHAT	0.369	0.358	0.386	0.328	0.332

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Max GIGACHAT	0.42	0.422	0.451	0.377	0.386

Model, team	Women	Men	LGBT	Nationalities	Migrants	Other
GigaChat Max GIGACHAT	0.824	0.543	0	0.135	0.286	0.77

GigaChat Max

Ratings for leaderboard tasks

Evaluation on open tasks:

Information about the submission:

Team:

Name of the ML model:

Link to the ML model:

Model type:

Architecture description:

Description of the training:

Pretrain data:

License:

Inference parameters

Ratings by subcategory

Confirm the deletion of the sub