Mistral-7B-Instruct-v0.1

MERA Created at 20.09.2024 09:22
0.369
The overall result
369
Place in the rating
Weak tasks:
90
RWSD
421
PARus
428
RCB
435
ruEthics
413
MultiQ
426
ruWorldTree
441
ruOpenBookQA
372
CheGeKa
426
ruMMLU
398
ruHateSpeech
456
ruDetox
397
ruHHH
474
ruTiE
117
ruHumanEval
481
USE
453
MathLogicQA
390
ruMultiAr
405
SimpleAr
316
LCS
376
BPS
294
ruModAr
389
MaMuRAMu
121
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.1 Accuracy
RCB 0.463 / 0.378 Accuracy F1 macro
USE 0.06 Grade norm
RWSD 0.612 Accuracy
PARus 0.614 Accuracy
ruTiE 0.511 Accuracy
MultiQ 0.259 / 0.112 F1 Exact match
CheGeKa 0.043 / 0.026 F1 Exact match
ruModAr 0.462 Exact match
MaMuRAMu 0.512 Accuracy
ruMultiAr 0.178 Exact match
ruCodeEval 0.17 / 0.33 / 0.402 Pass@k
MathLogicQA 0.315 Accuracy
ruWorldTree 0.669 / 0.668 Accuracy F1 macro
ruOpenBookQA 0.555 / 0.554 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 0.828 Accuracy
ruMMLU 0.415 Accuracy
SimpleAr 0.836 Exact match
ruHumanEval 0.178 / 0.315 / 0.354 Pass@k
ruHHH 0.545
ruHateSpeech 0.585
ruDetox 0.086
ruEthics
Correct God Ethical
Virtue 0.094 0.209 0.264
Law 0.118 0.22 0.28
Moral 0.127 0.215 0.279
Justice 0.078 0.182 0.247
Utilitarianism 0.101 0.167 0.223

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
bfloat16
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:

Mistral-7B-Instruct-v0.1

Model size

7.2B

Model type:

Opened

SFT

Additional links:

https://arxiv.org/abs/2310.06825

Architecture description:

"An instruct fine-tuned version of the Mistral-7B-v0.1, which 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:

Fine-tuned on instruction datasets publicly available on the Hugging Face repository. No proprietary data or training tricks were utilized.

Pretrain data:

Open Web data (further details are not disclosed by the authors).

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, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, rutie, bps, 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-Instruct-v0.1
MERA
0 0.1 0.467 0.067 0.033 0.1 0 - 0 0 0 0 0.1 0 0.1 0.133 0.067 0.033 0.033 0.033 0 0 0.033 0 0.067 0 0.2 0.033 0.067 0.133 0.1
Model, team Honest Helpful Harmless
Mistral-7B-Instruct-v0.1
MERA
0.557 0.576 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
Mistral-7B-Instruct-v0.1
MERA
0.385 0.349 0.447 0.598 0.487 0.562 0.563 0.466 0.441 0.43 0.254 0.389 0.32 0.5 0.512 0.46 0.41 0.389 0.3 0.511 0.25 0.456 0.36 0.474 0.214 0.46 0.307 0.407 0.514 0.35 0.47 0.661 0.454 0.657 0.475 0.363 0.28 0.503 0.338 0.335 0.571 0.397 0.421 0.345 0.496 0.444 0.52 0.233 0.323 0.38 0.42 0.544 0.387 0.479 0.44 0.515 0.523
Model, team SIM FL STA
Mistral-7B-Instruct-v0.1
MERA
0.802 0.509 0.256
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-Instruct-v0.1
MERA
0.489 0.475 0.4 0.435 0.368 0.414 0.431 0.561 0.462 0.446 0.564 0.467 0.342 0.512 0.439 0.494 0.607 0.511 0.263 0.667 0.421 0.593 0.467 0.503 0.533 0.591 0.513 0.456 0.772 0.6 0.733 0.615 0.482 0.596 0.515 0.554 0.378 0.689 0.333 0.538 0.567 0.444 0.667 0.289 0.828 0.644 0.638 0.364 0.585 0.632 0.622 0.565 0.544 0.481 0.419 0.468 0.567
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-Instruct-v0.1
MERA
0.094 0.118 0.127 0.078 0.101
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-Instruct-v0.1
MERA
0.209 0.22 0.215 0.182 0.167
Model, team Virtue Law Moral Justice Utilitarianism
Mistral-7B-Instruct-v0.1
MERA
0.264 0.28 0.279 0.247 0.223
Model, team Women Men LGBT Nationalities Migrants Other
Mistral-7B-Instruct-v0.1
MERA
0.546 0.657 0.588 0.595 0.571 0.607