Yi-1.5-34B-Chat

MERA Created at 20.09.2024 12:33
0.436
The overall result
273
Place in the rating
Weak tasks:
188
RWSD
304
PARus
143
RCB
254
ruEthics
292
MultiQ
316
ruWorldTree
388
ruOpenBookQA
361
CheGeKa
302
ruMMLU
337
ruHateSpeech
324
ruDetox
205
ruHHH
237
ruTiE
359
ruHumanEval
276
USE
194
MathLogicQA
151
ruMultiAr
221
SimpleAr
430
LCS
162
BPS
95
ruModAr
331
MaMuRAMu
285
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.082 Accuracy
RCB 0.557 / 0.501 Accuracy F1 macro
USE 0.147 Grade norm
RWSD 0.565 Accuracy
PARus 0.754 Accuracy
ruTiE 0.686 Accuracy
MultiQ 0.348 / 0.147 F1 Exact match
CheGeKa 0.048 / 0.022 F1 Exact match
ruModAr 0.704 Exact match
MaMuRAMu 0.577 Accuracy
ruMultiAr 0.319 Exact match
ruCodeEval 0.001 / 0.006 / 0.012 Pass@k
MathLogicQA 0.457 Accuracy
ruWorldTree 0.794 / 0.793 Accuracy F1 macro
ruOpenBookQA 0.635 / 0.632 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 0.965 Accuracy
ruMMLU 0.517 Accuracy
SimpleAr 0.97 Exact match
ruHumanEval 0.002 / 0.012 / 0.024 Pass@k
ruHHH 0.725
ruHateSpeech 0.638
ruDetox 0.146
ruEthics
Correct God Ethical
Virtue 0.256 0.379 0.256
Law 0.284 0.397 0.265
Moral 0.307 0.418 0.267
Justice 0.237 0.349 0.249
Utilitarianism 0.216 0.332 0.216

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
4 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Yi-1.5-34B-Chat

Model size

34.4B

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-1.5-34B-Chat
MERA
0.333 0.267 0.733 0.1 0 0.233 0 - 0.033 0 0.067 0.033 0.133 0.033 0.133 0.3 0.033 0.033 0.033 0.033 0.067 0.567 0.067 0 0.033 0.175 0.133 0.033 0.333 0.133 0.1
Model, team Honest Helpful Harmless
Yi-1.5-34B-Chat
MERA
0.721 0.678 0.776
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-1.5-34B-Chat
MERA
0.407 0.476 0.539 0.748 0.572 0.627 0.641 0.582 0.54 0.574 0.456 0.508 0.38 0.63 0.577 0.587 0.64 0.465 0.411 0.481 0.351 0.515 0.42 0.422 0.393 0.53 0.394 0.546 0.682 0.42 0.76 0.702 0.583 0.717 0.491 0.538 0.41 0.61 0.424 0.502 0.626 0.335 0.524 0.578 0.579 0.458 0.608 0.441 0.372 0.497 0.48 0.675 0.592 0.559 0.78 0.648 0.554
Model, team SIM FL STA
Yi-1.5-34B-Chat
MERA
0.432 0.539 0.742
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-1.5-34B-Chat
MERA
0.422 0.485 0.5 0.556 0.526 0.621 0.414 0.509 0.365 0.569 0.654 0.608 0.408 0.628 0.503 0.457 0.664 0.6 0.491 0.596 0.404 0.61 0.733 0.556 0.733 0.53 0.615 0.544 0.807 0.511 0.778 0.667 0.589 0.702 0.485 0.554 0.6 0.711 0.526 0.508 0.559 0.54 0.822 0.8 0.862 0.8 0.569 0.795 0.6 0.754 0.778 0.449 0.658 0.584 0.581 0.386 0.656
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat
MERA
0.256 0.284 0.307 0.237 0.216
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat
MERA
0.379 0.397 0.418 0.349 0.332
Model, team Virtue Law Moral Justice Utilitarianism
Yi-1.5-34B-Chat
MERA
0.256 0.265 0.267 0.249 0.216
Model, team Women Men LGBT Nationalities Migrants Other
Yi-1.5-34B-Chat
MERA
0.639 0.543 0.529 0.73 0.571 0.672