Yi-34B

MERA Created at 22.09.2024 22:10
0.44
The overall result
258
Place in the rating
Weak tasks:
231
RWSD
339
PARus
176
RCB
452
ruEthics
503
MultiQ
307
ruWorldTree
305
ruOpenBookQA
336
CheGeKa
276
ruMMLU
328
ruHateSpeech
452
ruDetox
361
ruHHH
275
ruTiE
155
ruHumanEval
269
USE
213
MathLogicQA
262
ruMultiAr
187
SimpleAr
360
LCS
415
BPS
105
ruModAr
301
MaMuRAMu
152
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.092 Accuracy
RCB 0.55 / 0.506 Accuracy F1 macro
USE 0.152 Grade norm
RWSD 0.55 Accuracy
PARus 0.708 Accuracy
ruTiE 0.658 Accuracy
MultiQ 0.184 / 0.114 F1 Exact match
CheGeKa 0.061 / 0.034 F1 Exact match
ruModAr 0.687 Exact match
MaMuRAMu 0.605 Accuracy
ruMultiAr 0.261 Exact match
ruCodeEval 0.091 / 0.231 / 0.305 Pass@k
MathLogicQA 0.44 Accuracy
ruWorldTree 0.806 / 0.804 Accuracy F1 macro
ruOpenBookQA 0.708 / 0.707 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 0.753 Accuracy
ruMMLU 0.535 Accuracy
SimpleAr 0.977 Exact match
ruHumanEval 0.112 / 0.263 / 0.335 Pass@k
ruHHH 0.579
ruHateSpeech 0.645
ruDetox 0.087
ruEthics
Correct God Ethical
Virtue 0.115 0.131 0.136
Law 0.103 0.137 0.11
Moral 0.135 0.139 0.118
Justice 0.096 0.125 0.139
Utilitarianism 0.066 0.098 0.083

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

Link to the ML model:

https://huggingface.co/01-ai/Yi-34B

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.

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-34B
MERA
0.467 0.233 0.633 0.1 0 0.167 0.033 - 0 0.067 0 0.033 0.1 0 0.1 0.267 0 0.033 0.067 0 0 0.4 0.033 0 0.1 0.308 0.2 0.133 0.167 0.167 0.167
Model, team Honest Helpful Harmless
Yi-34B
MERA
0.639 0.593 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
Yi-34B
MERA
0.422 0.482 0.618 0.795 0.611 0.736 0.68 0.601 0.522 0.578 0.439 0.46 0.37 0.657 0.584 0.613 0.61 0.528 0.289 0.504 0.343 0.544 0.29 0.416 0.446 0.49 0.437 0.565 0.661 0.37 0.68 0.785 0.644 0.717 0.498 0.509 0.38 0.639 0.43 0.493 0.702 0.478 0.531 0.507 0.622 0.532 0.735 0.359 0.433 0.511 0.49 0.709 0.577 0.567 0.77 0.727 0.606
Model, team SIM FL STA
Yi-34B
MERA
0.784 0.59 0.25
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
MERA
0.356 0.574 0.45 0.62 0.526 0.69 0.534 0.596 0.462 0.569 0.692 0.65 0.492 0.667 0.52 0.531 0.673 0.578 0.474 0.702 0.474 0.678 0.6 0.509 0.622 0.545 0.615 0.596 0.842 0.711 0.778 0.744 0.634 0.772 0.53 0.554 0.578 0.733 0.474 0.477 0.645 0.508 0.756 0.689 0.879 0.733 0.603 0.682 0.631 0.789 0.8 0.551 0.671 0.623 0.605 0.433 0.656
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B
MERA
0.115 0.103 0.135 0.096 0.066
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B
MERA
0.131 0.137 0.139 0.125 0.098
Model, team Virtue Law Moral Justice Utilitarianism
Yi-34B
MERA
0.136 0.11 0.118 0.139 0.083
Model, team Women Men LGBT Nationalities Migrants Other
Yi-34B
MERA
0.62 0.486 0.765 0.757 0.571 0.689