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
-
Butch
-
Transformers version
-
The number of GPUs and their type
-
Architecture
-

Team:

GIGACHAT

Name of the ML model:

GigaChat Lite

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

Expand information

Ratings by subcategory

Metric: Accuracy
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