Yi-1.5-34B-Chat-16K

MERA Created at 20.09.2024 07:46
0.444
The overall result
252
Place in the rating
Weak tasks:
160
RWSD
318
PARus
175
RCB
245
ruEthics
304
MultiQ
335
ruWorldTree
323
ruOpenBookQA
352
CheGeKa
282
ruMMLU
301
ruHateSpeech
208
ruDetox
264
ruHHH
194
ruTiE
222
USE
169
MathLogicQA
133
ruMultiAr
186
SimpleAr
359
LCS
140
BPS
108
ruModAr
311
MaMuRAMu
284
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.092 Accuracy
RCB 0.55 / 0.514 Accuracy F1 macro
USE 0.169 Grade norm
RWSD 0.573 Accuracy
PARus 0.736 Accuracy
ruTiE 0.718 Accuracy
MultiQ 0.338 / 0.184 F1 Exact match
CheGeKa 0.055 / 0.029 F1 Exact match
ruModAr 0.679 Exact match
MaMuRAMu 0.596 Accuracy
ruMultiAr 0.331 Exact match
ruCodeEval 0.001 / 0.006 / 0.012 Pass@k
MathLogicQA 0.469 Accuracy
ruWorldTree 0.775 / 0.774 Accuracy F1 macro
ruOpenBookQA 0.688 / 0.688 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 0.968 Accuracy
ruMMLU 0.529 Accuracy
SimpleAr 0.977 Exact match
ruHumanEval 0 / 0 / 0 Pass@k
ruHHH 0.663
ruHateSpeech 0.668
ruDetox 0.203
ruEthics
Correct God Ethical
Virtue 0.281 0.351 0.374
Law 0.285 0.337 0.364
Moral 0.312 0.369 0.407
Justice 0.245 0.31 0.357
Utilitarianism 0.227 0.342 0.332

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
4 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Yi-1.5-34B-Chat-16K

Model size

34.4B

Model type:

Opened

SFT

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 - 16384

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-Chat-16K
MERA
0.367 0.2 0.7 0.133 0.033 0.167 0.033 - 0 0.033 0.067 0.133 0.1 0.033 0.133 0.283 0 0 0.033 0 0.033 0.5 0.133 0.067 0.167 0.325 0.1 0.067 0.233 0.233 0.167
Model, team Honest Helpful Harmless
Yi-1.5-34B-Chat-16K
MERA
0.607 0.661 0.724
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-Chat-16K
MERA
0.393 0.476 0.572 0.701 0.592 0.677 0.68 0.579 0.525 0.538 0.447 0.548 0.41 0.685 0.577 0.624 0.66 0.521 0.411 0.565 0.324 0.556 0.38 0.468 0.33 0.53 0.405 0.574 0.706 0.4 0.69 0.719 0.558 0.687 0.547 0.513 0.43 0.61 0.404 0.542 0.616 0.397 0.517 0.642 0.594 0.519 0.637 0.504 0.415 0.472 0.47 0.717 0.59 0.597 0.81 0.673 0.601
Model, team SIM FL STA
Yi-1.5-34B-Chat-16K
MERA
0.502 0.581 0.771
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-Chat-16K
MERA
0.378 0.495 0.5 0.556 0.474 0.741 0.448 0.561 0.5 0.6 0.705 0.575 0.458 0.643 0.509 0.395 0.701 0.511 0.544 0.667 0.509 0.746 0.733 0.527 0.689 0.485 0.654 0.509 0.825 0.622 0.844 0.705 0.536 0.667 0.485 0.643 0.689 0.689 0.561 0.477 0.616 0.476 0.8 0.867 0.845 0.844 0.672 0.841 0.662 0.772 0.844 0.464 0.709 0.519 0.744 0.351 0.689
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat-16K
MERA
0.281 0.285 0.312 0.245 0.227
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat-16K
MERA
0.351 0.337 0.369 0.31 0.342
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat-16K
MERA
0.374 0.364 0.407 0.357 0.332
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-34B-Chat-16K
MERA
0.648 0.657 0.529 0.73 0.571 0.721