Qwen2-57B-A14B

MERA Created at 22.09.2024 21:57
0.493
The overall result
159
Place in the rating
Weak tasks:
214
RWSD
173
PARus
84
RCB
218
ruEthics
177
MultiQ
148
ruWorldTree
180
ruOpenBookQA
152
CheGeKa
145
ruMMLU
247
ruHateSpeech
132
ruDetox
199
ruHHH
136
ruTiE
217
ruHumanEval
238
USE
160
MathLogicQA
183
ruMultiAr
118
SimpleAr
524
LCS
327
BPS
163
ruModAr
143
MaMuRAMu
229
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.052 Accuracy
RCB 0.573 / 0.478 Accuracy F1 macro
USE 0.164 Grade norm
RWSD 0.554 Accuracy
PARus 0.87 Accuracy
ruTiE 0.776 Accuracy
MultiQ 0.444 / 0.287 F1 Exact match
CheGeKa 0.2 / 0.156 F1 Exact match
ruModAr 0.531 Exact match
MaMuRAMu 0.751 Accuracy
ruMultiAr 0.298 Exact match
ruCodeEval 0.012 / 0.052 / 0.091 Pass@k
MathLogicQA 0.473 Accuracy
ruWorldTree 0.949 / 0.949 Accuracy F1 macro
ruOpenBookQA 0.85 / 0.85 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 0.89 Accuracy
ruMMLU 0.658 Accuracy
SimpleAr 0.988 Exact match
ruHumanEval 0.042 / 0.152 / 0.232 Pass@k
ruHHH 0.73
ruHateSpeech 0.725
ruDetox 0.256
ruEthics
Correct God Ethical
Virtue 0.317 0.294 0.374
Law 0.324 0.318 0.351
Moral 0.344 0.307 0.392
Justice 0.289 0.254 0.313
Utilitarianism 0.298 0.29 0.327

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
bfloat16
Seed
1234
Batch
1
Transformers version
4.43.3
The number of GPUs and their type
8 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Qwen2-57B-A14B

Model size

57.4B

Model type:

Opened

Pretrain

Additional links:

https://qwenlm.github.io/blog/qwen2/

Architecture description:

Qwen 2 is a decoder-only language model.

Description of the training:

The model was pretrained with a large amount of data.

Pretrain data:

-

License:

Apache 2.0

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=["\n"];do_sample=false;until=["\n"]; \nchegeka - do_sample=false;until=["\n"];do_sample=false;until=["\n"]; \nrudetox - do_sample=false;until=["\n"];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, simplear, bps, lcs, chegeka, mathlogicqa, parus, rcb, multiq, rudetox, ruhatespeech, rummlu, rumodar, ruethics, ruworldtree, ruopenbookqa, ruhhh, rutie, ruhumaneval, rucodeeval - 131072

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
Qwen2-57B-A14B
MERA
0.467 0.333 0.733 0.067 0 0.3 0.033 - 0.033 0 0 0 0.167 0 0.1 0.167 0.033 0.033 0 0.033 0 0.567 0.133 0.033 0 0.35 0.033 0.167 0.233 0.1 0.233
Model, team Honest Helpful Harmless
Qwen2-57B-A14B
MERA
0.721 0.729 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
Qwen2-57B-A14B
MERA
0.63 0.494 0.763 0.863 0.745 0.826 0.825 0.727 0.713 0.677 0.526 0.484 0.46 0.759 0.797 0.699 0.69 0.826 0.522 0.725 0.334 0.789 0.34 0.688 0.545 0.81 0.483 0.685 0.751 0.54 0.67 0.793 0.65 0.848 0.732 0.714 0.45 0.832 0.47 0.631 0.854 0.732 0.621 0.589 0.82 0.606 0.843 0.459 0.5 0.669 0.56 0.827 0.754 0.857 0.77 0.824 0.798
Model, team SIM FL STA
Qwen2-57B-A14B
MERA
0.655 0.725 0.593
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
Qwen2-57B-A14B
MERA
0.467 0.861 0.7 0.667 0.829 0.759 0.672 0.702 0.731 0.738 0.769 0.708 0.558 0.783 0.667 0.679 0.71 0.8 0.579 0.842 0.596 0.881 0.867 0.787 0.711 0.758 0.756 0.702 0.895 0.756 0.822 0.859 0.705 0.877 0.667 0.696 0.667 0.822 0.632 0.631 0.849 0.905 0.8 0.911 0.914 0.844 0.879 0.841 0.8 0.93 0.8 0.754 0.772 0.714 0.512 0.637 0.833
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2-57B-A14B
MERA
0.317 0.324 0.344 0.289 0.298
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2-57B-A14B
MERA
0.294 0.318 0.307 0.254 0.29
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2-57B-A14B
MERA
0.374 0.351 0.392 0.313 0.327
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2-57B-A14B
MERA
0.722 0.629 0.824 0.757 0.429 0.77