Yi-1.5-34B

MERA Created at 20.09.2024 11:09
0.468
The overall result
200
Place in the rating
Weak tasks:
106
RWSD
287
PARus
358
RCB
283
ruEthics
390
MultiQ
340
ruWorldTree
343
ruOpenBookQA
334
CheGeKa
272
ruMMLU
277
ruHateSpeech
336
ruDetox
297
ruHHH
317
ruTiE
120
ruHumanEval
254
USE
129
MathLogicQA
128
ruMultiAr
164
SimpleAr
469
LCS
193
BPS
77
ruModAr
314
MaMuRAMu
87
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.072 Accuracy
RCB 0.498 / 0.46 Accuracy F1 macro
USE 0.157 Grade norm
RWSD 0.596 Accuracy
PARus 0.764 Accuracy
ruTiE 0.62 Accuracy
MultiQ 0.27 / 0.176 F1 Exact match
CheGeKa 0.066 / 0.036 F1 Exact match
ruModAr 0.729 Exact match
MaMuRAMu 0.593 Accuracy
ruMultiAr 0.334 Exact match
ruCodeEval 0.254 / 0.52 / 0.604 Pass@k
MathLogicQA 0.5 Accuracy
ruWorldTree 0.771 / 0.771 Accuracy F1 macro
ruOpenBookQA 0.675 / 0.673 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 0.957 Accuracy
ruMMLU 0.538 Accuracy
SimpleAr 0.981 Exact match
ruHumanEval 0.177 / 0.424 / 0.543 Pass@k
ruHHH 0.629
ruHateSpeech 0.687
ruDetox 0.142
ruEthics
Correct God Ethical
Virtue 0.257 0.377 0.277
Law 0.287 0.377 0.289
Moral 0.294 0.399 0.287
Justice 0.203 0.335 0.249
Utilitarianism 0.245 0.34 0.232

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-1.5-34B

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:

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-1.5-34B
MERA
0.467 0.233 0.7 0.133 0 0.167 0.033 - 0 0.067 0 0 0.1 0.033 0.133 0.183 0.033 0.033 0 0 0 0.567 0 0.067 0 0.35 0.133 0 0.233 0.3 0.133
Model, team Honest Helpful Harmless
Yi-1.5-34B
MERA
0.59 0.644 0.655
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-1.5-34B
MERA
0.385 0.464 0.592 0.731 0.601 0.692 0.709 0.621 0.494 0.534 0.404 0.532 0.35 0.685 0.576 0.575 0.62 0.451 0.433 0.534 0.363 0.503 0.4 0.451 0.438 0.57 0.422 0.574 0.698 0.4 0.72 0.769 0.601 0.758 0.54 0.551 0.48 0.642 0.457 0.532 0.652 0.368 0.538 0.602 0.631 0.583 0.627 0.478 0.422 0.5 0.54 0.65 0.623 0.609 0.81 0.667 0.58
Model, team SIM FL STA
Yi-1.5-34B
MERA
0.761 0.58 0.374
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-1.5-34B
MERA
0.311 0.515 0.467 0.546 0.539 0.741 0.448 0.579 0.365 0.523 0.718 0.617 0.442 0.651 0.52 0.494 0.682 0.622 0.544 0.702 0.526 0.729 0.756 0.462 0.644 0.5 0.654 0.491 0.772 0.667 0.8 0.718 0.58 0.702 0.5 0.661 0.644 0.689 0.544 0.477 0.608 0.444 0.822 0.822 0.81 0.822 0.621 0.841 0.662 0.737 0.778 0.478 0.696 0.584 0.628 0.363 0.711
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B
MERA
0.257 0.287 0.294 0.203 0.245
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B
MERA
0.377 0.377 0.399 0.335 0.34
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B
MERA
0.277 0.289 0.287 0.249 0.232
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-34B
MERA
0.713 0.771 0.706 0.676 0.286 0.639