Mistral-7B-v0.1

MERA Created at 20.09.2024 11:36
0.404
The overall result
320
Place in the rating
Weak tasks:
142
RWSD
458
PARus
422
RCB
418
ruEthics
438
MultiQ
347
ruWorldTree
325
ruOpenBookQA
199
CheGeKa
339
ruMMLU
389
ruHateSpeech
509
ruDetox
380
ruHHH
448
ruTiE
114
ruHumanEval
339
USE
423
MathLogicQA
359
ruMultiAr
268
SimpleAr
404
LCS
362
BPS
174
ruModAr
317
MaMuRAMu
106
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.086 Accuracy
RCB 0.466 / 0.403 Accuracy F1 macro
USE 0.107 Grade norm
RWSD 0.581 Accuracy
PARus 0.558 Accuracy
ruTiE 0.522 Accuracy
MultiQ 0.239 / 0.132 F1 Exact match
CheGeKa 0.165 / 0.118 F1 Exact match
ruModAr 0.517 Exact match
MaMuRAMu 0.585 Accuracy
ruMultiAr 0.195 Exact match
ruCodeEval 0.199 / 0.405 / 0.488 Pass@k
MathLogicQA 0.332 Accuracy
ruWorldTree 0.762 / 0.76 Accuracy F1 macro
ruOpenBookQA 0.688 / 0.684 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 0.849 Accuracy
ruMMLU 0.478 Accuracy
SimpleAr 0.948 Exact match
ruHumanEval 0.185 / 0.37 / 0.445 Pass@k
ruHHH 0.562
ruHateSpeech 0.592
ruDetox 0.068
ruEthics
Correct God Ethical
Virtue 0.094 0.201 0.202
Law 0.112 0.226 0.229
Moral 0.144 0.231 0.228
Justice 0.122 0.189 0.197
Utilitarianism 0.11 0.17 0.184

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
4.42.3
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Mistral-7B-v0.1

Model size

7.2B

Model type:

Opened

Pretrain

Additional links:

https://arxiv.org/abs/2310.06825

Architecture description:

Mistral-7B-v0.1 is a transformer model, with the following architecture choices: Grouped-Query Attention, Sliding-Window Attention, Byte-fallback BPE tokenizer. The context length is 8192.

Description of the training:

A pretrained generative text model.

Pretrain data:

-

License:

apache-2.0

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 - 32768

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
Mistral-7B-v0.1
MERA
0.033 0.3 0.667 0.2 0 0.1 0.033 - 0 0.067 0 0.033 0.2 0 0.133 0.15 0.067 0 0.067 0.033 0 0.167 0 0 0.1 0.108 0.1 0.067 0.2 0.133 0.2
Model, team Honest Helpful Harmless
Mistral-7B-v0.1
MERA
0.574 0.559 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
Mistral-7B-v0.1
MERA
0.489 0.44 0.487 0.692 0.536 0.692 0.544 0.579 0.543 0.516 0.316 0.357 0.33 0.574 0.584 0.578 0.54 0.444 0.333 0.649 0.257 0.713 0.32 0.532 0.321 0.53 0.361 0.519 0.576 0.39 0.69 0.719 0.497 0.707 0.509 0.419 0.31 0.581 0.344 0.438 0.631 0.401 0.524 0.379 0.569 0.347 0.662 0.296 0.369 0.48 0.41 0.62 0.415 0.462 0.57 0.648 0.549
Model, team SIM FL STA
Mistral-7B-v0.1
MERA
0.876 0.57 0.149
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
Mistral-7B-v0.1
MERA
0.444 0.564 0.467 0.528 0.579 0.569 0.517 0.579 0.577 0.538 0.603 0.525 0.433 0.628 0.544 0.506 0.617 0.667 0.333 0.702 0.316 0.729 0.556 0.58 0.689 0.591 0.577 0.579 0.842 0.689 0.733 0.667 0.464 0.614 0.439 0.571 0.356 0.711 0.456 0.585 0.673 0.556 0.822 0.422 0.862 0.733 0.793 0.545 0.523 0.789 0.667 0.652 0.696 0.532 0.558 0.515 0.678
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-v0.1
MERA
0.094 0.112 0.144 0.122 0.11
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-v0.1
MERA
0.201 0.226 0.231 0.189 0.17
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-v0.1
MERA
0.202 0.229 0.228 0.197 0.184
Model, team Women Men LGBT Nationalities Migrants Other
Mistral-7B-v0.1
MERA
0.602 0.629 0.529 0.73 0.429 0.508