gemma-2-27b

MERA Created at 22.09.2024 22:13
0.518
The overall result
127
Place in the rating
Weak tasks:
276
RWSD
109
PARus
213
RCB
187
ruEthics
602
MultiQ
164
ruWorldTree
149
ruOpenBookQA
81
CheGeKa
149
ruMMLU
194
ruHateSpeech
243
ruDetox
243
ruHHH
213
ruTiE
132
ruHumanEval
170
USE
228
MathLogicQA
108
ruMultiAr
123
SimpleAr
578
LCS
81
BPS
41
ruModAr
149
MaMuRAMu
126
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.03 Accuracy
RCB 0.541 / 0.47 Accuracy F1 macro
USE 0.201 Grade norm
RWSD 0.531 Accuracy
PARus 0.898 Accuracy
ruTiE 0.704 Accuracy
MultiQ 0.05 / 0.031 F1 Exact match
CheGeKa 0.352 / 0.281 F1 Exact match
ruModAr 0.87 Exact match
MaMuRAMu 0.746 Accuracy
ruMultiAr 0.344 Exact match
ruCodeEval 0.162 / 0.388 / 0.5 Pass@k
MathLogicQA 0.429 Accuracy
ruWorldTree 0.941 / 0.941 Accuracy F1 macro
ruOpenBookQA 0.868 / 0.868 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 0.984 Accuracy
ruMMLU 0.654 Accuracy
SimpleAr 0.987 Exact match
ruHumanEval 0.154 / 0.369 / 0.482 Pass@k
ruHHH 0.685
ruHateSpeech 0.762
ruDetox 0.183
ruEthics
Correct God Ethical
Virtue 0.294 0.381 0.419
Law 0.303 0.424 0.414
Moral 0.279 0.389 0.397
Justice 0.295 0.413 0.453
Utilitarianism 0.29 0.323 0.332

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
float32
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

Model size

27.2B

Model type:

Opened

Pretrain

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 tokens, the 9B model was trained with 8 trillion tokens, and 2B model was trained with 2 trillion tokens.

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
MERA
0.633 0.367 0.767 0.133 0.1 0.467 0.067 - 0 0 0 0 0.133 0 0.133 0.167 0.067 0 0 0 0.067 0.467 0.1 0.067 0.1 0.425 0.1 0.133 0.4 0.167 0.333
Model, team Honest Helpful Harmless
gemma-2-27b
MERA
0.656 0.661 0.741
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
MERA
0.615 0.518 0.75 0.846 0.768 0.806 0.806 0.695 0.716 0.691 0.57 0.492 0.43 0.713 0.746 0.717 0.7 0.819 0.489 0.74 0.301 0.778 0.32 0.671 0.491 0.73 0.525 0.639 0.747 0.52 0.79 0.793 0.712 0.869 0.694 0.697 0.45 0.861 0.483 0.616 0.808 0.706 0.648 0.509 0.884 0.653 0.863 0.411 0.415 0.67 0.58 0.835 0.751 0.811 0.77 0.764 0.865
Model, team SIM FL STA
gemma-2-27b
MERA
0.784 0.631 0.407
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
MERA
0.533 0.871 0.6 0.648 0.803 0.776 0.69 0.684 0.692 0.769 0.744 0.725 0.5 0.659 0.719 0.728 0.757 0.778 0.649 0.772 0.544 0.797 0.867 0.822 0.733 0.788 0.731 0.614 0.93 0.844 0.844 0.846 0.696 0.807 0.667 0.75 0.778 0.822 0.561 0.692 0.845 0.937 0.756 0.756 0.897 0.844 0.845 0.818 0.692 0.912 0.8 0.754 0.81 0.701 0.628 0.637 0.833
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b
MERA
0.294 0.303 0.279 0.295 0.29
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b
MERA
0.381 0.424 0.389 0.413 0.323
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-27b
MERA
0.419 0.414 0.397 0.453 0.332
Model, team Women Men LGBT Nationalities Migrants Other
gemma-2-27b
MERA
0.759 0.657 0.824 0.811 0.429 0.82