Meta-Llama-3.1-70B

MERA Created at 20.09.2024 17:09
0.571
The overall result
88
Place in the rating
Weak tasks:
66
RWSD
58
PARus
211
RCB
226
ruEthics
168
MultiQ
89
ruWorldTree
106
ruOpenBookQA
47
CheGeKa
103
ruMMLU
112
ruHateSpeech
295
ruDetox
189
ruHHH
129
ruTiE
74
ruHumanEval
131
USE
183
MathLogicQA
139
ruMultiAr
87
SimpleAr
90
LCS
134
BPS
55
ruModAr
94
MaMuRAMu
144
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.16 Accuracy
RCB 0.541 / 0.498 Accuracy F1 macro
USE 0.249 Grade norm
RWSD 0.638 Accuracy
PARus 0.924 Accuracy
ruTiE 0.779 Accuracy
MultiQ 0.45 / 0.293 F1 Exact match
CheGeKa 0.4 / 0.322 F1 Exact match
ruModAr 0.831 Exact match
MaMuRAMu 0.802 Accuracy
ruMultiAr 0.327 Exact match
ruCodeEval 0.107 / 0.306 / 0.415 Pass@k
MathLogicQA 0.462 Accuracy
ruWorldTree 0.971 / 0.971 Accuracy F1 macro
ruOpenBookQA 0.898 / 0.897 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 0.969 Accuracy
ruMMLU 0.707 Accuracy
SimpleAr 0.991 Exact match
ruHumanEval 0.271 / 0.542 / 0.64 Pass@k
ruHHH 0.736
ruHateSpeech 0.804
ruDetox 0.161
ruEthics
Correct God Ethical
Virtue 0.245 0.288 0.3
Law 0.272 0.305 0.309
Moral 0.304 0.308 0.314
Justice 0.296 0.278 0.301
Utilitarianism 0.249 0.293 0.292

Information about the submission

Mera version
v.1.2.0
Torch Version
2.3.1
The version of the codebase
6ec745e0
CUDA version
12.1
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.43.1
The number of GPUs and their type
8 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Meta-Llama-3.1-70B

Model size

70.6B

Model type:

Opened

Pretrain

Additional links:

https://huggingface.co/meta-llama/Meta-Llama-3.1-70B

Architecture description:

A pretrained 70B auto-regressive generative model with Grouped-Query Attention.

Description of the training:

-

Pretrain data:

Llama 3.1 was pretrained on ~15 trillion tokens of data from publicly available sources with a cutoff of December 2023.

License:

https://github.com/meta-llama/llama-models/blob/main/models/llama3_1/LICENSE

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 - 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
Meta-Llama-3.1-70B
MERA
0.433 0.4 0.867 0.1 0.067 0.367 0.033 - 0 0 0 0.033 0.333 0.033 0.1 0.317 0.067 0 0 0 0.1 0.667 0.233 0.167 0.2 0.525 0.167 0.2 0.433 0.367 0.367
Model, team Honest Helpful Harmless
Meta-Llama-3.1-70B
MERA
0.689 0.78 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
Meta-Llama-3.1-70B
MERA
0.689 0.518 0.822 0.838 0.84 0.861 0.816 0.807 0.858 0.744 0.614 0.54 0.5 0.843 0.84 0.772 0.81 0.868 0.5 0.786 0.378 0.877 0.4 0.671 0.554 0.8 0.547 0.657 0.796 0.47 0.81 0.884 0.724 0.879 0.743 0.756 0.52 0.868 0.556 0.714 0.869 0.835 0.676 0.56 0.884 0.648 0.912 0.444 0.511 0.752 0.64 0.895 0.749 0.811 0.81 0.861 0.902
Model, team SIM FL STA
Meta-Llama-3.1-70B
MERA
0.841 0.595 0.338
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
Meta-Llama-3.1-70B
MERA
0.644 0.871 0.667 0.667 0.882 0.862 0.707 0.702 0.904 0.769 0.756 0.758 0.625 0.76 0.819 0.765 0.71 0.822 0.684 0.877 0.702 0.881 0.889 0.811 0.822 0.879 0.782 0.737 0.93 0.844 0.889 0.846 0.759 0.895 0.773 0.839 0.778 0.867 0.649 0.769 0.898 0.937 0.844 0.844 0.914 0.867 0.862 0.75 0.8 0.93 0.8 0.942 0.81 0.727 0.651 0.778 0.889
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Meta-Llama-3.1-70B
MERA
0.245 0.272 0.304 0.296 0.249
Model, team Virtue Law Moral Justice Utilitarianism
Meta-Llama-3.1-70B
MERA
0.288 0.305 0.308 0.278 0.293
Model, team Virtue Law Moral Justice Utilitarianism
Meta-Llama-3.1-70B
MERA
0.3 0.309 0.314 0.301 0.292
Model, team Women Men LGBT Nationalities Migrants Other
Meta-Llama-3.1-70B
MERA
0.815 0.714 0.824 0.811 0.714 0.836