gemma-2-2b

MERA Created at 22.09.2024 22:15
0.314
The overall result
460
Place in the rating
Weak tasks:
408
RWSD
532
PARus
546
RCB
588
MultiQ
516
ruWorldTree
504
ruOpenBookQA
348
CheGeKa
468
ruMMLU
472
ruHateSpeech
537
ruDetox
474
ruHHH
467
ruTiE
137
ruHumanEval
550
USE
482
MathLogicQA
351
ruMultiAr
362
SimpleAr
435
LCS
491
BPS
194
ruModAr
491
MaMuRAMu
136
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.082 Accuracy
RCB 0.349 / 0.252 Accuracy F1 macro
USE 0.041 Grade norm
RWSD 0.5 Accuracy
PARus 0.5 Accuracy
ruTiE 0.513 Accuracy
MultiQ 0.088 / 0.037 F1 Exact match
CheGeKa 0.057 / 0.041 F1 Exact match
ruModAr 0.51 Exact match
MaMuRAMu 0.421 Accuracy
ruMultiAr 0.202 Exact match
ruCodeEval 0.126 / 0.281 / 0.354 Pass@k
MathLogicQA 0.305 Accuracy
ruWorldTree 0.51 / 0.503 Accuracy F1 macro
ruOpenBookQA 0.473 / 0.468 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 0.565 Accuracy
ruMMLU 0.378 Accuracy
SimpleAr 0.896 Exact match
ruHumanEval 0.143 / 0.282 / 0.341 Pass@k
ruHHH 0.5
ruHateSpeech 0.532
ruDetox 0.059
ruEthics
Correct God Ethical
Virtue -0.011 0.066 0.017
Law -0.028 0.026 0.004
Moral -0.011 0.027 0.006
Justice -0.031 0.053 0.001
Utilitarianism 0.004 0.046 -0.018

Information about the submission

Mera version
v.1.2.0
Torch Version
2.3.1
The version of the codebase
aec92f8
CUDA version
12.1
Precision of the model weights
float32
Seed
1234
Batch
1
Transformers version
4.43.3
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

gemma-2-2b

Model size

2.6B

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. 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, rutie, ruhumaneval, rucodeeval - 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-2-2b
MERA
0 0.033 0.267 0.133 0.033 0.2 0 - 0 0 0 0 0 0 0.133 0.1 0.033 0 0.033 0.033 0 0 0 0 0.033 0.025 0.067 0.033 0 0 0.067
Model, team Honest Helpful Harmless
gemma-2-2b
MERA
0.525 0.475 0.5
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-2-2b
MERA
0.341 0.392 0.349 0.611 0.474 0.527 0.447 0.424 0.466 0.448 0.228 0.27 0.36 0.426 0.416 0.436 0.36 0.292 0.289 0.435 0.279 0.368 0.31 0.364 0.321 0.36 0.318 0.38 0.522 0.27 0.46 0.587 0.38 0.606 0.366 0.355 0.31 0.442 0.338 0.35 0.449 0.279 0.421 0.35 0.412 0.338 0.422 0.237 0.319 0.361 0.35 0.443 0.351 0.353 0.43 0.418 0.347
Model, team SIM FL STA
gemma-2-2b
MERA
0.817 0.576 0.175
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-2-2b
MERA
0.289 0.416 0.383 0.352 0.382 0.397 0.345 0.316 0.269 0.415 0.372 0.325 0.342 0.488 0.31 0.37 0.514 0.533 0.368 0.561 0.228 0.508 0.356 0.456 0.4 0.424 0.423 0.439 0.702 0.467 0.667 0.474 0.393 0.474 0.485 0.429 0.289 0.511 0.316 0.4 0.404 0.413 0.467 0.4 0.759 0.667 0.293 0.364 0.431 0.632 0.511 0.319 0.519 0.416 0.256 0.386 0.544
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-2b
MERA
-0.011 -0.028 -0.011 -0.031 0.004
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-2b
MERA
0.066 0.026 0.027 0.053 0.046
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-2b
MERA
0.017 0.004 0.006 0.001 -0.018
Model, team Women Men LGBT Nationalities Migrants Other
gemma-2-2b
MERA
0.519 0.714 0.471 0.622 0.429 0.426