Yi-1.5-6B

MERA Created at 20.09.2024 12:58
0.358
The overall result
379
Place in the rating
Weak tasks:
372
RWSD
472
PARus
458
RCB
511
MultiQ
487
ruWorldTree
477
ruOpenBookQA
533
CheGeKa
449
ruMMLU
551
ruHateSpeech
542
ruDetox
399
ruHHH
438
ruTiE
143
ruHumanEval
468
USE
343
MathLogicQA
175
ruMultiAr
230
SimpleAr
420
LCS
236
BPS
245
ruModAr
463
MaMuRAMu
95
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.084 Accuracy
RCB 0.427 / 0.421 Accuracy F1 macro
USE 0.064 Grade norm
RWSD 0.508 Accuracy
PARus 0.544 Accuracy
ruTiE 0.528 Accuracy
MultiQ 0.181 / 0.104 F1 Exact match
CheGeKa 0.009 / 0.002 F1 Exact match
ruModAr 0.482 Exact match
MaMuRAMu 0.439 Accuracy
ruMultiAr 0.301 Exact match
ruCodeEval 0.218 / 0.461 / 0.549 Pass@k
MathLogicQA 0.366 Accuracy
ruWorldTree 0.566 / 0.557 Accuracy F1 macro
ruOpenBookQA 0.513 / 0.502 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 0.945 Accuracy
ruMMLU 0.397 Accuracy
SimpleAr 0.968 Exact match
ruHumanEval 0.135 / 0.331 / 0.421 Pass@k
ruHHH 0.545
ruHateSpeech 0.483
ruDetox 0.057
ruEthics
Correct God Ethical
Virtue 0.05 0.046 -0.003
Law 0.085 0.059 0.056
Moral 0.064 0.07 0.029
Justice 0.066 0.041 0.018
Utilitarianism 0.035 0.057 -0.017

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:

Yi-1.5-6B

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:

Yi-1.5 is an upgraded version of Yi (which was trained on 3T multilingual corpus). It is continuously pre-trained on Yi with a high-quality corpus of 500B tokens and fine-tuned on 3M diverse fine-tuning samples.

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-1.5-6B
MERA
0.167 0.033 0.4 0.1 0 0.1 0 - 0 0 0 0 0.1 0 0.1 0.1 0.033 0 0.033 0.033 0 0.133 0 0 0 0.05 0.067 0.133 0.133 0.1 0.1
Model, team Honest Helpful Harmless
Yi-1.5-6B
MERA
0.557 0.61 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-1.5-6B
MERA
0.252 0.392 0.441 0.603 0.422 0.478 0.495 0.463 0.383 0.426 0.219 0.373 0.34 0.5 0.411 0.457 0.49 0.285 0.244 0.45 0.261 0.404 0.29 0.358 0.268 0.5 0.332 0.537 0.555 0.37 0.54 0.669 0.374 0.576 0.408 0.406 0.33 0.429 0.305 0.36 0.465 0.342 0.497 0.406 0.38 0.384 0.456 0.359 0.34 0.364 0.42 0.46 0.418 0.483 0.68 0.461 0.404
Model, team SIM FL STA
Yi-1.5-6B
MERA
0.813 0.573 0.177
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-1.5-6B
MERA
0.378 0.426 0.467 0.472 0.342 0.448 0.362 0.333 0.519 0.415 0.5 0.492 0.275 0.512 0.339 0.272 0.57 0.289 0.263 0.491 0.298 0.407 0.4 0.414 0.533 0.5 0.551 0.386 0.684 0.4 0.689 0.372 0.482 0.439 0.303 0.446 0.467 0.467 0.368 0.354 0.388 0.365 0.711 0.6 0.724 0.733 0.31 0.682 0.508 0.614 0.622 0.319 0.544 0.403 0.465 0.327 0.444
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B
MERA
0.05 0.085 0.064 0.066 0.035
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B
MERA
0.046 0.059 0.07 0.041 0.057
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B
MERA
-0.003 0.056 0.029 0.018 -0.017
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-6B
MERA
0.435 0.457 0.353 0.622 0.571 0.525