Yi-6B-Chat

MERA Created at 20.09.2024 10:34
0.227
The overall result
559
Place in the rating
Weak tasks:
417
RWSD
522
PARus
573
RCB
442
ruEthics
571
MultiQ
558
ruWorldTree
548
ruOpenBookQA
543
CheGeKa
542
ruMMLU
415
ruHateSpeech
578
ruDetox
471
ruHHH
505
ruTiE
578
USE
564
MathLogicQA
595
SimpleAr
354
LCS
565
BPS
531
MaMuRAMu
+15
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.094 Accuracy
RCB 0.34 / 0.327 Accuracy F1 macro
USE 0.028 Grade norm
RWSD 0.5 Accuracy
PARus 0.504 Accuracy
ruTiE 0.506 Accuracy
MultiQ 0.116 / 0.018 F1 Exact match
CheGeKa 0.009 / 0 F1 Exact match
ruModAr 0 Exact match
MaMuRAMu 0.355 Accuracy
ruMultiAr 0 Exact match
ruCodeEval 0 / 0 / 0 Pass@k
MathLogicQA 0.256 Accuracy
ruWorldTree 0.387 / 0.37 Accuracy F1 macro
ruOpenBookQA 0.385 / 0.377 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 0.495 Accuracy
ruMMLU 0.316 Accuracy
SimpleAr 0.002 Exact match
ruHumanEval 0 / 0 / 0 Pass@k
ruHHH 0.506
ruHateSpeech 0.566
ruDetox 0.052
ruEthics
Correct God Ethical
Virtue 0.129 0.124 0.174
Law 0.103 0.097 0.163
Moral 0.16 0.127 0.179
Justice 0.112 0.09 0.117
Utilitarianism 0.088 0.073 0.122

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-Chat

Model size

6.1B

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:

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. The released chat model has undergone exclusive training using Supervised Fine-Tuning (SFT).

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

Description of the template:
{% if not add_generation_prompt is defined %}{% set add_generation_prompt = false %}{% endif %}{% for message in messages %}{{'<|im_start|>' + message['role'] + ' \n' + message['content'] + '<|im_end|>' + ' \n'}}{% endfor %}{% if add_generation_prompt %}{{ '<|im_start|>assistant \n' }}{% endif %}

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-Chat
MERA
0 0.033 0.367 0.033 0 0.033 0 - 0 0.033 0 0 0 0 0.033 0.1 0 0.033 0 0 0 0 0 0 0 0.05 0 0 0 0 0
Model, team Honest Helpful Harmless
Yi-6B-Chat
MERA
0.492 0.542 0.483
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-Chat
MERA
0.319 0.271 0.362 0.466 0.301 0.368 0.311 0.338 0.321 0.341 0.254 0.27 0.22 0.352 0.317 0.416 0.36 0.215 0.256 0.351 0.257 0.292 0.21 0.341 0.268 0.38 0.282 0.37 0.367 0.31 0.42 0.512 0.264 0.333 0.325 0.295 0.29 0.374 0.272 0.286 0.369 0.301 0.421 0.271 0.311 0.352 0.319 0.215 0.27 0.321 0.36 0.388 0.282 0.336 0.46 0.412 0.264
Model, team SIM FL STA
Yi-6B-Chat
MERA
0.24 0.512 0.606
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-Chat
MERA
0.267 0.436 0.283 0.417 0.368 0.345 0.362 0.263 0.327 0.292 0.436 0.333 0.233 0.349 0.287 0.358 0.449 0.333 0.246 0.456 0.298 0.356 0.333 0.32 0.333 0.273 0.308 0.439 0.649 0.333 0.511 0.436 0.438 0.456 0.242 0.304 0.244 0.289 0.281 0.385 0.343 0.381 0.444 0.333 0.483 0.333 0.31 0.227 0.431 0.491 0.356 0.29 0.405 0.325 0.326 0.292 0.478
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-Chat
MERA
0.129 0.103 0.16 0.112 0.088
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-Chat
MERA
0.124 0.097 0.127 0.09 0.073
Model, team Virtue Law Moral Justice Utilitarianism
Yi-6B-Chat
MERA
0.174 0.163 0.179 0.117 0.122
Model, team Women Men LGBT Nationalities Migrants Other
Yi-6B-Chat
MERA
0.574 0.571 0.588 0.595 0.429 0.541