GigaChat Lite

GIGACHAT Created at 15.07.2024 06:04

0.504

The overall result

The submission does not contain all the required tasks

Ratings for leaderboard tasks

The table will scroll to the left

Task name	Result	Metric
LCS	0.084	Accuracy
RCB	0.543 / 0.452	Accuracy F1 macro
USE	0.284	Grade norm
RWSD	0.627	Accuracy
PARus	0.848	Accuracy
ruTiE	0.726	Accuracy
MultiQ	0.193 / 0.071	F1 Exact match
CheGeKa	0.063 / 0	F1 Exact match
ruModAr	0.77	Exact match
ruMultiAr	0.216	Exact match
MathLogicQA	0.45	Accuracy
ruWorldTree	0.897 / 0.897	Accuracy F1 macro
ruOpenBookQA	0.823 / 0.822	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

0.412

Accuracy

ruMMLU

0.783

Accuracy

SimpleAr

0.9

Exact match

ruHumanEval

0.018 / 0.088 / 0.177

Pass@k

ruHHH

0.753

ruHateSpeech

0.774

ruDetox

0.05

ruEthics

	Correct	God	Ethical
Virtue	-0.336	-0.294	-0.314
Law	-0.332	-0.3	-0.301
Moral	-0.351	-0.305	-0.323
Justice	-0.31	-0.261	-0.273
Utilitarianism	-0.237	-0.201	-0.242

Information about the submission

Mera version

Torch Version

The version of the codebase

CUDA version

Precision of the model weights

Seed

Batch

Transformers version

The number of GPUs and their type

Architecture

Team:

GIGACHAT

Name of the ML model:

GigaChat Lite

Link to the ML model:

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

Additional links:

https://developers.sber.ru/docs/ru/gigachat/api/overview

Architecture description:

GigaChat Lite (version `GigaChat:4.0.26.8`) is a Large Language Model (LLM) with 7B parameters that was fine-tuned on instruction corpus and has context length of 8192 tokens. The version is available for users via API since 13.07.

Description of the training:

Pretrain data:

Training Details:

License:

Strategy, generation and parameters:

Code version v.1.1.0. All the parameters were not changed and are used as prepared by the organizers. Details: - 2 x NVIDIA A100 + accelerate - dtype float16 - Pytorch 2.3.1 + CUDA 12.1 - Transformers 4.42.3 - Context length 8192

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 Lite GIGACHAT	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-	-

Model, team	Honest	Helpful	Harmless
GigaChat Lite GIGACHAT	0.721	0.729	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 Lite GIGACHAT	0.7	0.813	0.6	0.657	0.857	1	0.8	0.647	0.8	0.8	0.818	0.8	0.4	0.615	0.636	0.7	0.6	0.815	0.9	1	0.3	0.808	1	0.804	0.6	0.909	0.75	0.786	1	0.909	0.7	0.778	0.7	1	1	0.9	0.9	0.81	0.7	0.8	0.937	1	0.7	0.7	0.938	0.8	1	0.6	0.7	1	0.591	0.938	0.882	0.8	0.583	0.576	0.741

Model, team	SIM	FL	STA
GigaChat Lite GIGACHAT	0.307	0.821	0.147

Coorect

Good

Ethical

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Lite GIGACHAT	-0.336	-0.332	-0.351	-0.31	-0.237

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Lite GIGACHAT	-0.294	-0.3	-0.305	-0.261	-0.201

Model, team	Virtue	Law	Moral	Justice	Utilitarianism
GigaChat Lite GIGACHAT	-0.314	-0.301	-0.323	-0.273	-0.242

Model, team	Women	Men	LGBT	Nationalities	Migrants	Other
GigaChat Lite GIGACHAT	0.759	0.8	0.706	0.73	0.429	0.869

GigaChat Lite

Ratings for leaderboard tasks

Evaluation on open tasks:

Information about the submission

Team:

Name of the ML model:

Link to the ML model:

Additional links:

Architecture description:

Description of the training:

Pretrain data:

Training Details:

License:

Strategy, generation and parameters:

Ratings by subcategory

Confirm the deletion of the sub