GigaChat

GIGACHAT Created at 01.11.2024 08:35

0.5

The overall result

107

Place in the rating

In the top by tasks:

ruModAr

The task is one of the main ones

Weak tasks:

505

RWSD

202

PARus

265

RCB

256

ruEthics

206

MultiQ

181

ruWorldTree

177

ruOpenBookQA

CheGeKa

199

ruMMLU

324

ruHateSpeech

280

ruDetox

124

ruHHH

132

ruTiE

157

ruHumanEval

USE

106

MathLogicQA

276

ruMultiAr

245

SimpleAr

180

LCS

313

BPS

125

MaMuRAMu

126

ruCodeEval

+18

Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name	Result	Metric
LCS	0.116	Accuracy
RCB	0.509 / 0.314	Accuracy F1 macro
USE	0.34	Grade norm
RWSD	0.427	Accuracy
PARus	0.82	Accuracy
ruTiE	0.728	Accuracy
MultiQ	0.367 / 0.25	F1 Exact match
CheGeKa	0.307 / 0.26	F1 Exact match
ruModAr	0.863	Exact match
MaMuRAMu	0.724	Accuracy
ruMultiAr	0.225	Exact match
ruCodeEval	0.077 / 0.151 / 0.183	Pass@k
MathLogicQA	0.473	Accuracy
ruWorldTree	0.91 / 0.91	Accuracy F1 macro
ruOpenBookQA	0.815 / 0.654	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

0.839

Accuracy

ruMMLU

0.566

Accuracy

SimpleAr

0.938

Exact match

ruHumanEval

0.063 / 0.11 / 0.134

Pass@k

ruHHH

0.758

ruHateSpeech

0.604

ruDetox

0.143

ruEthics

	Correct	God	Ethical
Virtue	0.212	0.285	0.283
Law	0.256	0.29	0.281
Moral	0.246	0.294	0.3
Justice	0.187	0.249	0.262
Utilitarianism	0.195	0.244	0.287

Information about the submission:

Mera version

v.1.2.0

Torch Version

2.4.0

The version of the codebase

db539c9

CUDA version

12.1

Precision of the model weights

Seed

1234

Butch

Transformers version

4.46.0.dev0

The number of GPUs and their type

5 x NVIDIA H100 80GB HBM3

Architecture

gigachat_llms

Team:

GIGACHAT

Name of the ML model:

GigaChat

Link to the ML model:

https://developers.sber.ru/portal/products/gigachat-api

Model type:

Closed

API

SFT

MoE

Architecture description:

GigaChat (version GigaChat:26.20) is a Large Language Model (LLM) that was fine-tuned on instruction corpus and has context length of 32k tokens. GigaChat Pro is Mixture of Experts model and has 3B active parameters. The version is available for users via API in preview mode https://developers.sber.ru/portal/products/gigachat-api

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"];temperature=0.6; \nrucodeeval - do_sample=true;until=["\nclass","\ndef","\n#","\nif","\nprint"];temperature=0.6;

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.

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
GigaChat GIGACHAT	0.833	0.333	0.633	0.167	0.133	0.333	0.167	-	0.2	0.2	0.267	0.1	0.433	0.1	0.067	0.533	0.067	0	0.3	0.067	0.1	0.467	0.3	0.033	0.1	0.692	0.433	0.3	0.5	0.467	0.633

Model, team	Honest	Helpful	Harmless
GigaChat GIGACHAT	0.689	0.78	0.81

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 GIGACHAT	0.548	0.482	0.717	0.808	0.65	0.806	0.699	0.666	0.648	0.578	0.421	0.405	0.38	0.63	0.727	0.584	0.62	0.632	0.322	0.695	0.277	0.778	0.31	0.555	0.402	0.65	0.397	0.63	0.661	0.39	0.71	0.793	0.607	0.747	0.623	0.521	0.31	0.71	0.391	0.498	0.793	0.533	0.586	0.451	0.768	0.454	0.691	0.341	0.411	0.567	0.52	0.717	0.615	0.689	0.69	0.679	0.751

Model, team	SIM	FL	STA
GigaChat GIGACHAT	0.273	0.736	0.735

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
GigaChat GIGACHAT	0.467	0.802	0.567	0.648	0.789	0.793	0.621	0.702	0.788	0.677	0.667	0.667	0.475	0.829	0.731	0.704	0.72	0.667	0.526	0.807	0.333	0.763	0.689	0.746	0.689	0.803	0.795	0.737	0.86	0.756	0.822	0.833	0.714	0.825	0.682	0.679	0.667	0.8	0.509	0.585	0.829	0.857	0.778	0.622	0.914	0.822	0.845	0.591	0.831	0.86	0.822	0.754	0.759	0.649	0.465	0.737	0.8

Coorect

Good

Ethical

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat GIGACHAT	0.212	0.256	0.246	0.187	0.195

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat GIGACHAT	0.285	0.29	0.294	0.249	0.244

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat GIGACHAT	0.283	0.281	0.3	0.262	0.287

Model, team	Women	Men	LGBT	Nationalities	Migrants	Other
GigaChat GIGACHAT	0.778	0.6	0	0.162	0.286	0.77

GigaChat

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