Yi-34B-200K

MERA Created at 22.09.2024 22:09
0.426
The overall result
287
Place in the rating
Weak tasks:
340
RWSD
332
PARus
244
RCB
252
ruEthics
505
MultiQ
339
ruWorldTree
309
ruOpenBookQA
444
CheGeKa
286
ruMMLU
305
ruHateSpeech
422
ruDetox
316
ruHHH
289
ruTiE
127
ruHumanEval
321
USE
218
MathLogicQA
266
ruMultiAr
254
SimpleAr
464
LCS
502
BPS
97
ruModAr
308
MaMuRAMu
139
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.074 Accuracy
RCB 0.534 / 0.438 Accuracy F1 macro
USE 0.118 Grade norm
RWSD 0.515 Accuracy
PARus 0.716 Accuracy
ruTiE 0.648 Accuracy
MultiQ 0.184 / 0.014 F1 Exact match
CheGeKa 0.022 / 0.007 F1 Exact match
ruModAr 0.702 Exact match
MaMuRAMu 0.6 Accuracy
ruMultiAr 0.259 Exact match
ruCodeEval 0.117 / 0.283 / 0.348 Pass@k
MathLogicQA 0.435 Accuracy
ruWorldTree 0.773 / 0.769 Accuracy F1 macro
ruOpenBookQA 0.7 / 0.699 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 0.55 Accuracy
ruMMLU 0.527 Accuracy
SimpleAr 0.957 Exact match
ruHumanEval 0.165 / 0.312 / 0.36 Pass@k
ruHHH 0.618
ruHateSpeech 0.664
ruDetox 0.102
ruEthics
Correct God Ethical
Virtue 0.263 0.308 0.278
Law 0.259 0.303 0.262
Moral 0.286 0.334 0.298
Justice 0.217 0.257 0.244
Utilitarianism 0.244 0.249 0.275

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
4.43.3
The number of GPUs and their type
4 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Yi-34B-200K

Model size

34.4B

Model type:

Opened

Pretrain

Architecture description:

The Yi series models adopt the same model architecture as Llama but are NOT derivatives of Llama. Context window is extended to 200K.

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 - 200000

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-34B-200K
MERA
0.133 0.1 0.6 0.2 0.033 0.1 0 - 0 0.067 0 0 0 0 0.1 0.167 0.033 0 0.033 0.033 0 0.2 0 0 0.033 0.167 0.167 0.1 0.4 0.333 0.333
Model, team Honest Helpful Harmless
Yi-34B-200K
MERA
0.557 0.678 0.621
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-34B-200K
MERA
0.385 0.476 0.559 0.739 0.595 0.766 0.67 0.617 0.531 0.574 0.395 0.444 0.4 0.667 0.582 0.618 0.6 0.458 0.289 0.55 0.307 0.573 0.33 0.491 0.384 0.53 0.441 0.583 0.673 0.38 0.7 0.793 0.564 0.758 0.536 0.521 0.41 0.623 0.391 0.453 0.672 0.445 0.552 0.517 0.601 0.523 0.657 0.344 0.415 0.484 0.46 0.675 0.587 0.588 0.69 0.648 0.653
Model, team SIM FL STA
Yi-34B-200K
MERA
0.734 0.549 0.326
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-34B-200K
MERA
0.4 0.554 0.433 0.574 0.566 0.69 0.517 0.579 0.481 0.646 0.731 0.65 0.45 0.659 0.491 0.543 0.607 0.556 0.509 0.684 0.351 0.746 0.689 0.538 0.644 0.515 0.628 0.614 0.86 0.733 0.778 0.756 0.625 0.772 0.545 0.589 0.511 0.733 0.491 0.446 0.624 0.524 0.711 0.756 0.828 0.778 0.621 0.636 0.569 0.719 0.756 0.522 0.696 0.597 0.558 0.456 0.644
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B-200K
MERA
0.263 0.259 0.286 0.217 0.244
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B-200K
MERA
0.308 0.303 0.334 0.257 0.249
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B-200K
MERA
0.278 0.262 0.298 0.244 0.275
Model, team Women Men LGBT Nationalities Migrants Other
Yi-34B-200K
MERA
0.648 0.629 0.706 0.676 0.286 0.738