Yi-6B

MERA Created at 20.09.2024 08:25
0.317
The overall result
457
Place in the rating
Weak tasks:
394
RWSD
447
PARus
488
RCB
583
MultiQ
511
ruWorldTree
474
ruOpenBookQA
485
CheGeKa
455
ruMMLU
493
ruHateSpeech
598
ruDetox
513
ruHHH
478
ruTiE
171
ruHumanEval
430
USE
379
MathLogicQA
394
ruMultiAr
286
SimpleAr
352
LCS
534
BPS
339
ruModAr
482
MaMuRAMu
184
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.094 Accuracy
RCB 0.404 / 0.364 Accuracy F1 macro
USE 0.071 Grade norm
RWSD 0.504 Accuracy
PARus 0.574 Accuracy
ruTiE 0.51 Accuracy
MultiQ 0.104 / 0.064 F1 Exact match
CheGeKa 0.015 / 0.002 F1 Exact match
ruModAr 0.407 Exact match
MaMuRAMu 0.43 Accuracy
ruMultiAr 0.171 Exact match
ruCodeEval 0.055 / 0.135 / 0.195 Pass@k
MathLogicQA 0.354 Accuracy
ruWorldTree 0.524 / 0.522 Accuracy F1 macro
ruOpenBookQA 0.515 / 0.515 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 0.513 Accuracy
ruMMLU 0.389 Accuracy
SimpleAr 0.94 Exact match
ruHumanEval 0.091 / 0.172 / 0.213 Pass@k
ruHHH 0.478
ruHateSpeech 0.521
ruDetox 0.045
ruEthics
Correct God Ethical
Virtue -0.014 0.11 0.072
Law -0.013 0.071 0.079
Moral -0.014 0.104 0.068
Justice -0.012 0.075 0.083
Utilitarianism -0.016 0.087 0.058

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
float16
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:

Yi-6B

Link to the ML model:

https://huggingface.co/01-ai/Yi-6B

Model size

6.1B

Model type:

Opened

Pretrain

Architecture description:

The Yi series models adopt the same model architecture as Llama but are NOT derivatives of Llama.

Description of the training:

Yi has independently created its own high-quality training datasets, efficient training pipelines, and robust training infrastructure entirely from the ground up.

Pretrain data:

Targeted as a bilingual language model and trained on 3T multilingual corpus with 4K sequence length and can be extended to 32K during inference time. Training data date up to June 2023.

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 - 4096

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
Yi-6B
MERA
0.267 0.067 0.533 0.033 0 0.1 0 - 0 0 0 0.033 0.133 0 0.133 0.117 0.033 0.033 0.033 0.033 0 0.033 0 0 0.067 0.033 0.133 0.1 0.033 0.067 0.167
Model, team Honest Helpful Harmless
Yi-6B
MERA
0.492 0.475 0.466
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
Yi-6B
MERA
0.281 0.355 0.428 0.641 0.405 0.512 0.456 0.482 0.364 0.341 0.228 0.397 0.31 0.481 0.382 0.462 0.44 0.306 0.267 0.382 0.248 0.357 0.2 0.341 0.295 0.47 0.352 0.426 0.58 0.3 0.53 0.678 0.429 0.535 0.396 0.359 0.38 0.439 0.291 0.325 0.48 0.419 0.483 0.358 0.421 0.356 0.412 0.3 0.355 0.359 0.39 0.468 0.374 0.387 0.52 0.521 0.399
Model, team SIM FL STA
Yi-6B
MERA
0.844 0.55 0.125
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
Yi-6B
MERA
0.422 0.446 0.367 0.444 0.395 0.448 0.431 0.421 0.327 0.415 0.372 0.442 0.292 0.473 0.316 0.296 0.542 0.378 0.193 0.404 0.316 0.508 0.4 0.373 0.378 0.47 0.474 0.491 0.737 0.467 0.644 0.449 0.411 0.474 0.455 0.321 0.378 0.6 0.263 0.292 0.351 0.397 0.533 0.467 0.69 0.622 0.362 0.318 0.631 0.596 0.689 0.319 0.532 0.584 0.512 0.38 0.556
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B
MERA
-0.014 -0.013 -0.014 -0.012 -0.016
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B
MERA
0.11 0.071 0.104 0.075 0.087
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B
MERA
0.072 0.079 0.068 0.083 0.058
Model, team Women Men LGBT Nationalities Migrants Other
Yi-6B
MERA
0.546 0.657 0.353 0.568 0.143 0.459