Yi-1.5-9B

MERA Created at 20.09.2024 13:48
0.43
The overall result
280
Place in the rating
Weak tasks:
173
RWSD
314
PARus
319
RCB
353
ruEthics
425
MultiQ
363
ruWorldTree
389
ruOpenBookQA
442
CheGeKa
350
ruMMLU
274
ruHateSpeech
487
ruDetox
348
ruHHH
319
ruTiE
90
ruHumanEval
323
USE
250
MathLogicQA
221
ruMultiAr
199
SimpleAr
441
LCS
204
BPS
181
ruModAr
388
MaMuRAMu
71
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.08 Accuracy
RCB 0.511 / 0.484 Accuracy F1 macro
USE 0.117 Grade norm
RWSD 0.569 Accuracy
PARus 0.74 Accuracy
ruTiE 0.618 Accuracy
MultiQ 0.253 / 0.18 F1 Exact match
CheGeKa 0.022 / 0.01 F1 Exact match
ruModAr 0.514 Exact match
MaMuRAMu 0.512 Accuracy
ruMultiAr 0.28 Exact match
ruCodeEval 0.293 / 0.552 / 0.634 Pass@k
MathLogicQA 0.419 Accuracy
ruWorldTree 0.749 / 0.748 Accuracy F1 macro
ruOpenBookQA 0.635 / 0.633 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 0.47 Accuracy
SimpleAr 0.976 Exact match
ruHumanEval 0.234 / 0.491 / 0.604 Pass@k
ruHHH 0.59
ruHateSpeech 0.691
ruDetox 0.073
ruEthics
Correct God Ethical
Virtue 0.191 0.283 0.158
Law 0.221 0.29 0.151
Moral 0.226 0.303 0.154
Justice 0.191 0.285 0.15
Utilitarianism 0.157 0.242 0.142

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-1.5-9B

Model size

8.8B

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-1.5-9B
MERA
0.3 0.2 0.567 0.1 0 0.033 0 - 0 0 0 0 0.333 0 0.1 0.183 0.033 0.033 0.033 0 0 0.333 0.033 0 0 0.158 0.133 0.1 0.2 0.1 0.333
Model, team Honest Helpful Harmless
Yi-1.5-9B
MERA
0.557 0.576 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-1.5-9B
MERA
0.304 0.434 0.507 0.722 0.523 0.642 0.563 0.518 0.506 0.453 0.404 0.421 0.31 0.574 0.48 0.526 0.57 0.396 0.333 0.489 0.287 0.409 0.33 0.376 0.402 0.44 0.398 0.556 0.62 0.31 0.67 0.711 0.571 0.636 0.419 0.479 0.35 0.539 0.404 0.433 0.566 0.397 0.545 0.528 0.513 0.505 0.578 0.43 0.376 0.428 0.47 0.599 0.487 0.529 0.68 0.594 0.492
Model, team SIM FL STA
Yi-1.5-9B
MERA
0.851 0.557 0.185
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-1.5-9B
MERA
0.333 0.455 0.417 0.565 0.434 0.552 0.448 0.404 0.327 0.492 0.615 0.625 0.325 0.519 0.368 0.321 0.598 0.467 0.421 0.649 0.281 0.508 0.622 0.467 0.667 0.485 0.564 0.491 0.754 0.422 0.733 0.462 0.554 0.509 0.439 0.625 0.578 0.578 0.456 0.431 0.457 0.381 0.778 0.733 0.845 0.756 0.379 0.773 0.569 0.667 0.756 0.449 0.684 0.545 0.674 0.374 0.556
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B
MERA
0.191 0.221 0.226 0.191 0.157
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B
MERA
0.283 0.29 0.303 0.285 0.242
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B
MERA
0.158 0.151 0.154 0.15 0.142
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-9B
MERA
0.685 0.657 0.706 0.73 0.286 0.738