Yi-1.5-6B-Chat

MERA Created at 20.09.2024 13:03
0.334
The overall result
428
Place in the rating
Weak tasks:
249
RWSD
470
PARus
472
RCB
444
ruEthics
419
MultiQ
485
ruWorldTree
488
ruOpenBookQA
446
CheGeKa
474
ruMMLU
400
ruHateSpeech
469
ruDetox
346
ruHHH
417
ruTiE
365
ruHumanEval
386
USE
414
MathLogicQA
283
ruMultiAr
270
SimpleAr
365
LCS
207
BPS
250
ruModAr
500
MaMuRAMu
234
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.092 Accuracy
RCB 0.418 / 0.402 Accuracy F1 macro
USE 0.082 Grade norm
RWSD 0.546 Accuracy
PARus 0.546 Accuracy
ruTiE 0.544 Accuracy
MultiQ 0.257 / 0.127 F1 Exact match
CheGeKa 0.022 / 0.005 F1 Exact match
ruModAr 0.48 Exact match
MaMuRAMu 0.415 Accuracy
ruMultiAr 0.252 Exact match
ruCodeEval 0.008 / 0.028 / 0.043 Pass@k
MathLogicQA 0.337 Accuracy
ruWorldTree 0.571 / 0.567 Accuracy F1 macro
ruOpenBookQA 0.5 / 0.495 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 0.956 Accuracy
ruMMLU 0.377 Accuracy
SimpleAr 0.947 Exact match
ruHumanEval 0.002 / 0.009 / 0.018 Pass@k
ruHHH 0.596
ruHateSpeech 0.577
ruDetox 0.081
ruEthics
Correct God Ethical
Virtue 0.222 0.268 0.094
Law 0.22 0.296 0.075
Moral 0.237 0.272 0.121
Justice 0.184 0.262 0.072
Utilitarianism 0.173 0.207 0.114

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-Chat

Model size

6.1B

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 - 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-Chat
MERA
0.167 0.033 0.4 0.067 0 0.1 0 - 0 0.033 0 0 0.067 0 0.067 0.283 0 0 0.133 0.033 0.033 0.067 0 0.033 0.067 0.067 0.067 0.067 0.267 0.1 0.167
Model, team Honest Helpful Harmless
Yi-1.5-6B-Chat
MERA
0.557 0.593 0.638
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-Chat
MERA
0.267 0.361 0.428 0.59 0.386 0.517 0.476 0.473 0.414 0.444 0.298 0.381 0.33 0.481 0.395 0.422 0.48 0.313 0.267 0.427 0.278 0.304 0.22 0.295 0.321 0.41 0.304 0.417 0.531 0.35 0.46 0.554 0.429 0.455 0.366 0.393 0.38 0.397 0.305 0.345 0.47 0.246 0.448 0.432 0.358 0.352 0.417 0.344 0.337 0.357 0.32 0.405 0.359 0.437 0.65 0.424 0.29
Model, team SIM FL STA
Yi-1.5-6B-Chat
MERA
0.448 0.4 0.574
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-Chat
MERA
0.267 0.366 0.317 0.37 0.329 0.448 0.345 0.316 0.327 0.477 0.551 0.433 0.25 0.457 0.333 0.284 0.57 0.333 0.439 0.491 0.246 0.322 0.422 0.337 0.556 0.333 0.5 0.439 0.684 0.4 0.733 0.295 0.295 0.351 0.333 0.429 0.444 0.489 0.404 0.369 0.38 0.444 0.667 0.489 0.655 0.8 0.31 0.682 0.569 0.632 0.689 0.188 0.595 0.468 0.465 0.275 0.478
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B-Chat
MERA
0.222 0.22 0.237 0.184 0.173
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B-Chat
MERA
0.268 0.296 0.272 0.262 0.207
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-6B-Chat
MERA
0.094 0.075 0.121 0.072 0.114
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-6B-Chat
MERA
0.583 0.543 0.529 0.622 0.714 0.557