Yi-6B-200K

MERA Created at 20.09.2024 09:41
0.306
The overall result
475
Place in the rating
Weak tasks:
475
RWSD
467
PARus
450
RCB
562
MultiQ
522
ruWorldTree
478
ruOpenBookQA
537
CheGeKa
456
ruMMLU
573
ruHateSpeech
599
ruDetox
457
ruHHH
470
ruTiE
181
ruHumanEval
497
USE
364
MathLogicQA
428
ruMultiAr
357
SimpleAr
409
LCS
510
BPS
385
ruModAr
498
MaMuRAMu
179
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.086 Accuracy
RCB 0.438 / 0.413 Accuracy F1 macro
USE 0.057 Grade norm
RWSD 0.485 Accuracy
PARus 0.548 Accuracy
ruTiE 0.512 Accuracy
MultiQ 0.125 / 0.048 F1 Exact match
CheGeKa 0.009 / 0.002 F1 Exact match
ruModAr 0.367 Exact match
MaMuRAMu 0.416 Accuracy
ruMultiAr 0.144 Exact match
ruCodeEval 0.057 / 0.136 / 0.189 Pass@k
MathLogicQA 0.359 Accuracy
ruWorldTree 0.469 / 0.456 Accuracy F1 macro
ruOpenBookQA 0.513 / 0.508 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 0.545 Accuracy
ruMMLU 0.389 Accuracy
SimpleAr 0.902 Exact match
ruHumanEval 0.086 / 0.175 / 0.226 Pass@k
ruHHH 0.511
ruHateSpeech 0.468
ruDetox 0.045
ruEthics
Correct God Ethical
Virtue 0 -0.055 0.022
Law 0 -0.051 0.002
Moral 0 -0.057 0.045
Justice 0 -0.049 0.038
Utilitarianism 0 -0.065 0.026

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
float16
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-6B-200K

Model size

6.1B

Model type:

Opened

Pretrain

Architecture description:

The Yi series models adopt the same model architecture as Llama but are NOT derivatives of Llama. This release contains model with the same parameter size as the previous release, except that the 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-6B-200K
MERA
0.133 0 0.433 0 0 0.067 0 - 0 0 0 0 0.133 0 0.067 0.017 0.033 0.033 0.067 0.033 0 0 0 0 0.1 0.017 0.033 0.2 0.167 0.067 0.267
Model, team Honest Helpful Harmless
Yi-6B-200K
MERA
0.508 0.508 0.517
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-6B-200K
MERA
0.326 0.343 0.414 0.598 0.444 0.552 0.495 0.463 0.383 0.417 0.289 0.294 0.36 0.481 0.41 0.442 0.47 0.319 0.222 0.397 0.273 0.363 0.32 0.364 0.259 0.43 0.336 0.481 0.543 0.32 0.52 0.645 0.429 0.545 0.366 0.385 0.38 0.429 0.358 0.305 0.52 0.371 0.421 0.363 0.388 0.296 0.426 0.27 0.351 0.382 0.39 0.46 0.372 0.353 0.53 0.436 0.44
Model, team SIM FL STA
Yi-6B-200K
MERA
0.848 0.541 0.121
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-6B-200K
MERA
0.333 0.426 0.317 0.537 0.355 0.345 0.466 0.333 0.385 0.431 0.423 0.392 0.383 0.403 0.31 0.284 0.551 0.289 0.333 0.456 0.281 0.39 0.422 0.391 0.511 0.455 0.538 0.351 0.684 0.333 0.756 0.487 0.393 0.351 0.394 0.375 0.333 0.489 0.298 0.338 0.327 0.317 0.578 0.422 0.741 0.6 0.379 0.364 0.554 0.614 0.711 0.29 0.481 0.519 0.419 0.287 0.522
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-200K
MERA
0 0 0 0 0
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-200K
MERA
-0.055 -0.051 -0.057 -0.049 -0.065
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-200K
MERA
0.022 0.002 0.045 0.038 0.026
Model, team Women Men LGBT Nationalities Migrants Other
Yi-6B-200K
MERA
0.472 0.4 0.471 0.432 0.571 0.508