Yi-1.5-9B-32K

MERA Created at 20.09.2024 13:51
0.428
The overall result
283
Place in the rating
Weak tasks:
174
RWSD
294
PARus
352
RCB
404
ruEthics
482
MultiQ
374
ruWorldTree
380
ruOpenBookQA
477
CheGeKa
345
ruMMLU
285
ruHateSpeech
515
ruDetox
309
ruHHH
322
ruTiE
88
ruHumanEval
317
USE
269
MathLogicQA
206
ruMultiAr
190
SimpleAr
431
LCS
139
BPS
176
ruModAr
385
MaMuRAMu
67
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.082 Accuracy
RCB 0.5 / 0.483 Accuracy F1 macro
USE 0.12 Grade norm
RWSD 0.569 Accuracy
PARus 0.76 Accuracy
ruTiE 0.614 Accuracy
MultiQ 0.204 / 0.083 F1 Exact match
CheGeKa 0.016 / 0.002 F1 Exact match
ruModAr 0.516 Exact match
MaMuRAMu 0.516 Accuracy
ruMultiAr 0.288 Exact match
ruCodeEval 0.302 / 0.586 / 0.677 Pass@k
MathLogicQA 0.413 Accuracy
ruWorldTree 0.739 / 0.738 Accuracy F1 macro
ruOpenBookQA 0.64 / 0.638 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 0.969 Accuracy
ruMMLU 0.474 Accuracy
SimpleAr 0.977 Exact match
ruHumanEval 0.248 / 0.483 / 0.585 Pass@k
ruHHH 0.624
ruHateSpeech 0.679
ruDetox 0.067
ruEthics
Correct God Ethical
Virtue 0.175 0.224 0.107
Law 0.223 0.229 0.117
Moral 0.194 0.209 0.116
Justice 0.158 0.218 0.115
Utilitarianism 0.16 0.178 0.1

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-32K

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 - 32768

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-32K
MERA
0.233 0.133 0.633 0.167 0 0.067 0 - 0.033 0 0 0 0.133 0 0.133 0.167 0.033 0.033 0.033 0 0 0.333 0.033 0 0.067 0.175 0.133 0.1 0.267 0.233 0.233
Model, team Honest Helpful Harmless
Yi-1.5-9B-32K
MERA
0.59 0.661 0.621
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-32K
MERA
0.348 0.422 0.507 0.709 0.539 0.657 0.563 0.54 0.478 0.534 0.386 0.437 0.35 0.583 0.512 0.54 0.61 0.361 0.3 0.466 0.287 0.433 0.31 0.382 0.375 0.47 0.392 0.565 0.608 0.31 0.65 0.694 0.558 0.667 0.423 0.47 0.39 0.539 0.404 0.424 0.596 0.386 0.552 0.552 0.507 0.472 0.578 0.404 0.401 0.428 0.42 0.62 0.492 0.538 0.75 0.582 0.497
Model, team SIM FL STA
Yi-1.5-9B-32K
MERA
0.859 0.567 0.155
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-32K
MERA
0.333 0.436 0.5 0.574 0.395 0.534 0.448 0.316 0.442 0.538 0.641 0.6 0.308 0.558 0.392 0.321 0.617 0.422 0.404 0.649 0.351 0.542 0.556 0.479 0.6 0.485 0.615 0.526 0.825 0.444 0.756 0.538 0.571 0.456 0.439 0.571 0.511 0.556 0.491 0.415 0.469 0.365 0.711 0.733 0.81 0.822 0.379 0.773 0.585 0.632 0.733 0.449 0.671 0.545 0.581 0.398 0.533
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-32K
MERA
0.175 0.223 0.194 0.158 0.16
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-32K
MERA
0.224 0.229 0.209 0.218 0.178
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-9B-32K
MERA
0.107 0.117 0.116 0.115 0.1
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-9B-32K
MERA
0.657 0.6 0.706 0.784 0.571 0.705