gemma-2-27b-it

MERA Created at 22.09.2024 22:11
0.579
The overall result
79
Place in the rating
Weak tasks:
181
RWSD
69
PARus
107
RCB
233
ruEthics
142
MultiQ
103
ruWorldTree
77
ruOpenBookQA
84
CheGeKa
124
ruMMLU
46
ruHateSpeech
183
ruDetox
71
ruHHH
92
ruTiE
62
ruHumanEval
109
USE
124
MathLogicQA
111
ruMultiAr
78
SimpleAr
517
LCS
72
BPS
68
ruModAr
122
MaMuRAMu
81
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.054 Accuracy
RCB 0.566 / 0.514 Accuracy F1 macro
USE 0.29 Grade norm
RWSD 0.565 Accuracy
PARus 0.92 Accuracy
ruTiE 0.821 Accuracy
MultiQ 0.461 / 0.342 F1 Exact match
CheGeKa 0.351 / 0.288 F1 Exact match
ruModAr 0.769 Exact match
MaMuRAMu 0.768 Accuracy
ruMultiAr 0.342 Exact match
ruCodeEval 0.259 / 0.586 / 0.689 Pass@k
MathLogicQA 0.506 Accuracy
ruWorldTree 0.966 / 0.966 Accuracy F1 macro
ruOpenBookQA 0.915 / 0.915 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 0.986 Accuracy
ruMMLU 0.672 Accuracy
SimpleAr 0.993 Exact match
ruHumanEval 0.321 / 0.615 / 0.683 Pass@k
ruHHH 0.854
ruHateSpeech 0.853
ruDetox 0.213
ruEthics
Correct God Ethical
Virtue 0.329 0.287 0.35
Law 0.32 0.284 0.324
Moral 0.345 0.29 0.365
Justice 0.284 0.24 0.312
Utilitarianism 0.293 0.255 0.28

Information about the submission

Mera version
v.1.2.0
Torch Version
2.4.0
The version of the codebase
6ec745e0
CUDA version
12.1
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.43.2
The number of GPUs and their type
4 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

gemma-2-27b-it

Model size

27.2B

Model type:

Opened

SFT

Architecture description:

Gemma is built from the same research and technology used to create the Gemini models. They are text-to-text, decoder-only large language models, available in English, with open weights for both pre-trained variants and instruction-tuned variants

Description of the training:

-

Pretrain data:

These models were trained on a dataset of text data that includes a wide variety of sources. The 27B model was trained with 13 trillion tokenss.

License:

gemma

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, bps, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, ruhumaneval, rucodeeval, rutie - 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
gemma-2-27b-it
MERA
0.667 0.533 0.767 0.2 0.167 0.467 0.1 - 0 0 0.033 0.067 0.267 0.033 0.133 0.5 0 0.033 0.067 0.033 0.1 0.8 0.433 0.367 0.067 0.5 0.133 0.2 0.433 0.433 0.333
Model, team Honest Helpful Harmless
gemma-2-27b-it
MERA
0.836 0.814 0.914
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
gemma-2-27b-it
MERA
0.622 0.53 0.757 0.85 0.771 0.811 0.786 0.714 0.747 0.744 0.588 0.595 0.41 0.778 0.768 0.725 0.69 0.861 0.511 0.771 0.318 0.813 0.4 0.682 0.545 0.83 0.509 0.657 0.776 0.5 0.71 0.835 0.73 0.859 0.721 0.714 0.47 0.897 0.596 0.67 0.828 0.75 0.655 0.536 0.89 0.639 0.853 0.452 0.486 0.697 0.61 0.848 0.754 0.811 0.75 0.83 0.87
Model, team SIM FL STA
gemma-2-27b-it
MERA
0.357 0.789 0.78
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
gemma-2-27b-it
MERA
0.533 0.871 0.667 0.667 0.803 0.741 0.724 0.702 0.712 0.8 0.756 0.725 0.575 0.752 0.719 0.741 0.766 0.8 0.684 0.842 0.561 0.814 0.889 0.834 0.733 0.803 0.718 0.667 0.93 0.822 0.844 0.833 0.75 0.842 0.667 0.804 0.844 0.844 0.632 0.754 0.837 0.873 0.844 0.822 0.879 0.822 0.914 0.864 0.815 0.912 0.822 0.812 0.81 0.701 0.651 0.673 0.844
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b-it
MERA
0.329 0.32 0.345 0.284 0.293
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b-it
MERA
0.287 0.284 0.29 0.24 0.255
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b-it
MERA
0.35 0.324 0.365 0.312 0.28
Model, team Women Men LGBT Nationalities Migrants Other
gemma-2-27b-it
MERA
0.861 0.771 1 0.811 0.571 0.902