gemma-2-9b

MERA Created at 22.09.2024 22:17
0.453
The overall result
235
Place in the rating
Weak tasks:
203
RWSD
191
PARus
376
RCB
467
ruEthics
603
MultiQ
247
ruWorldTree
272
ruOpenBookQA
137
CheGeKa
225
ruMMLU
375
ruHateSpeech
410
ruDetox
354
ruHHH
326
ruTiE
135
ruHumanEval
264
USE
302
MathLogicQA
204
ruMultiAr
167
SimpleAr
483
LCS
194
BPS
144
ruModAr
236
MaMuRAMu
141
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.068 Accuracy
RCB 0.489 / 0.399 Accuracy F1 macro
USE 0.154 Grade norm
RWSD 0.558 Accuracy
PARus 0.854 Accuracy
ruTiE 0.609 Accuracy
MultiQ 0.049 / 0.029 F1 Exact match
CheGeKa 0.22 / 0.166 F1 Exact match
ruModAr 0.592 Exact match
MaMuRAMu 0.689 Accuracy
ruMultiAr 0.288 Exact match
ruCodeEval 0.112 / 0.321 / 0.421 Pass@k
MathLogicQA 0.391 Accuracy
ruWorldTree 0.88 / 0.88 Accuracy F1 macro
ruOpenBookQA 0.755 / 0.751 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 0.957 Accuracy
ruMMLU 0.591 Accuracy
SimpleAr 0.981 Exact match
ruHumanEval 0.145 / 0.309 / 0.402 Pass@k
ruHHH 0.584
ruHateSpeech 0.608
ruDetox 0.108
ruEthics
Correct God Ethical
Virtue 0.103 0.085 0.056
Law 0.16 0.107 0.063
Moral 0.135 0.102 0.066
Justice 0.123 0.098 0.047
Utilitarianism 0.093 0.096 0.046

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
1 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

gemma-2-9b

Model size

9.2B

Model type:

Opened

Pretrain

Additional links:

https://storage.googleapis.com/deepmind-media/gemma/gemma-2-report.pdf

Architecture description:

Gemma 2 is a text-to-text, decoder-only large language model, available in English. • SentencePiece tokenizer of size 256k

Description of the training:

Pretraining Gemma 2: • Knowledge Distillation • The model was trained with 8 trillion tokens • TPUv4

Pretrain data:

Pretraining data: Web Documents, Code, Mathematics

License:

https://ai.google.dev/gemma/terms

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, 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-2-9b
MERA
0.367 0.167 0.7 0.067 0.033 0.3 0.033 - 0 0.033 0 0 0.2 0.033 0.133 0.083 0.033 0 0.033 0.033 0 0.5 0 0.033 0.1 0.4 0.067 0.167 0.167 0.1 0.167
Model, team Honest Helpful Harmless
gemma-2-9b
MERA
0.574 0.627 0.552
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-2-9b
MERA
0.578 0.488 0.711 0.812 0.696 0.701 0.709 0.659 0.63 0.655 0.474 0.397 0.4 0.741 0.668 0.656 0.68 0.757 0.433 0.656 0.266 0.713 0.33 0.624 0.455 0.71 0.463 0.583 0.722 0.49 0.69 0.793 0.663 0.808 0.649 0.637 0.35 0.8 0.43 0.571 0.702 0.588 0.538 0.493 0.787 0.546 0.745 0.356 0.45 0.58 0.38 0.785 0.651 0.681 0.75 0.745 0.736
Model, team SIM FL STA
gemma-2-9b
MERA
0.826 0.619 0.239
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-2-9b
MERA
0.533 0.762 0.617 0.63 0.697 0.776 0.552 0.737 0.635 0.708 0.731 0.633 0.492 0.674 0.585 0.568 0.692 0.733 0.561 0.789 0.368 0.729 0.667 0.757 0.6 0.758 0.692 0.649 0.877 0.756 0.8 0.769 0.589 0.86 0.545 0.732 0.689 0.822 0.614 0.677 0.796 0.81 0.733 0.8 0.862 0.8 0.81 0.818 0.631 0.807 0.689 0.652 0.823 0.662 0.488 0.591 0.756
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-9b
MERA
0.103 0.16 0.135 0.123 0.093
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-9b
MERA
0.085 0.107 0.102 0.098 0.096
Model, team Virtue Law Moral Justice Utilitarianism
gemma-2-9b
MERA
0.056 0.063 0.066 0.047 0.046
Model, team Women Men LGBT Nationalities Migrants Other
gemma-2-9b
MERA
0.62 0.486 0.529 0.757 0.286 0.623