Yi-1.5-9B-Chat

MERA Created at 22.09.2024 22:06
0.393
The overall result
341
Place in the rating
Weak tasks:
178
RWSD
392
PARus
347
RCB
304
ruEthics
440
MultiQ
390
ruWorldTree
353
ruOpenBookQA
483
CheGeKa
353
ruMMLU
271
ruHateSpeech
418
ruDetox
349
ruHHH
277
ruTiE
386
ruHumanEval
280
USE
243
MathLogicQA
219
ruMultiAr
148
SimpleAr
324
LCS
249
BPS
187
ruModAr
414
MaMuRAMu
268
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.098 Accuracy
RCB 0.502 / 0.457 Accuracy F1 macro
USE 0.141 Grade norm
RWSD 0.569 Accuracy
PARus 0.65 Accuracy
ruTiE 0.656 Accuracy
MultiQ 0.238 / 0.132 F1 Exact match
CheGeKa 0.015 / 0.002 F1 Exact match
ruModAr 0.512 Exact match
MaMuRAMu 0.498 Accuracy
ruMultiAr 0.281 Exact match
ruCodeEval 0.002 / 0.012 / 0.024 Pass@k
MathLogicQA 0.423 Accuracy
ruWorldTree 0.718 / 0.718 Accuracy F1 macro
ruOpenBookQA 0.663 / 0.663 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 0.939 Accuracy
ruMMLU 0.465 Accuracy
SimpleAr 0.983 Exact match
ruHumanEval 0.001 / 0.006 / 0.012 Pass@k
ruHHH 0.59
ruHateSpeech 0.694
ruDetox 0.103
ruEthics
Correct God Ethical
Virtue 0.245 0.364 0.215
Law 0.296 0.352 0.233
Moral 0.282 0.385 0.239
Justice 0.267 0.35 0.19
Utilitarianism 0.237 0.316 0.201

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-Chat

Model size

8.8B

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-1.5-9B-Chat
MERA
0.367 0.133 0.6 0.2 0 0.133 0.033 - 0.033 0 0 0 0.033 0 0.067 0.417 0 0 0.033 0 0.033 0.3 0.1 0 0.133 0.183 0.133 0.1 0.4 0.2 0.2
Model, team Honest Helpful Harmless
Yi-1.5-9B-Chat
MERA
0.459 0.61 0.707
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-Chat
MERA
0.363 0.428 0.461 0.722 0.546 0.587 0.544 0.508 0.475 0.475 0.368 0.452 0.32 0.519 0.513 0.517 0.63 0.368 0.367 0.443 0.258 0.427 0.31 0.341 0.491 0.48 0.376 0.463 0.612 0.37 0.66 0.686 0.534 0.636 0.385 0.5 0.49 0.494 0.358 0.512 0.566 0.335 0.517 0.515 0.543 0.458 0.505 0.381 0.372 0.464 0.5 0.57 0.508 0.58 0.72 0.539 0.466
Model, team SIM FL STA
Yi-1.5-9B-Chat
MERA
0.486 0.467 0.58
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-Chat
MERA
0.378 0.426 0.4 0.528 0.368 0.603 0.431 0.404 0.308 0.4 0.577 0.517 0.408 0.481 0.357 0.383 0.589 0.467 0.404 0.526 0.281 0.475 0.689 0.402 0.622 0.455 0.577 0.404 0.737 0.511 0.689 0.526 0.509 0.544 0.379 0.679 0.511 0.533 0.333 0.492 0.502 0.349 0.711 0.8 0.707 0.8 0.397 0.795 0.585 0.719 0.756 0.464 0.658 0.558 0.558 0.316 0.578
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat
MERA
0.245 0.296 0.282 0.267 0.237
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat
MERA
0.364 0.352 0.385 0.35 0.316
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-Chat
MERA
0.215 0.233 0.239 0.19 0.201
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-9B-Chat
MERA
0.713 0.571 0.647 0.73 0.286 0.77