GPT3.5-turbo

MERA Created at 23.09.2024 14:23
0.438
The overall result
264
Place in the rating
Weak tasks:
337
RWSD
307
PARus
324
RCB
181
ruEthics
212
MultiQ
256
ruWorldTree
281
ruOpenBookQA
120
CheGeKa
273
ruMMLU
387
ruHateSpeech
116
ruDetox
215
ruHHH
294
ruTiE
266
ruHumanEval
282
USE
361
MathLogicQA
215
ruMultiAr
244
SimpleAr
163
LCS
214
BPS
219
ruModAr
248
MaMuRAMu
233
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.136 Accuracy
RCB 0.509 / 0.482 Accuracy F1 macro
USE 0.14 Grade norm
RWSD 0.515 Accuracy
PARus 0.752 Accuracy
ruTiE 0.644 Accuracy
MultiQ 0.406 / 0.223 F1 Exact match
CheGeKa 0.313 / 0.238 F1 Exact match
ruModAr 0.495 Exact match
MaMuRAMu 0.681 Accuracy
ruMultiAr 0.286 Exact match
ruCodeEval 0.009 / 0.02 / 0.03 Pass@k
MathLogicQA 0.36 Accuracy
ruWorldTree 0.874 / 0.7 Accuracy F1 macro
ruOpenBookQA 0.743 / 0.597 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 0.953 Accuracy
ruMMLU 0.538 Accuracy
SimpleAr 0.962 Exact match
ruHumanEval 0.01 / 0.015 / 0.018 Pass@k
ruHHH 0.719
ruHateSpeech 0.596
ruDetox 0.274
ruEthics
Correct God Ethical
Virtue 0.349 0.35 0.374
Law 0.328 0.331 0.343
Moral 0.362 0.361 0.386
Justice 0.296 0.289 0.324
Utilitarianism 0.281 0.287 0.311

Information about the submission

Mera version
v.1.2.0
Torch Version
2.0.1
The version of the codebase
7a72a611
CUDA version
11.7
Precision of the model weights
-
Seed
1234
Batch
1
Transformers version
4.36.2
The number of GPUs and their type
-
Architecture
openai-chat-completions

Team:

MERA

Name of the ML model:

GPT3.5-turbo

Link to the ML model:

https://openai.com/api/pricing/

Model type:

Closed

API

SFT

Architecture description:

Version: gpt-3.5-turbo-0125 Updated GPT 3.5 Turbo. The latest GPT-3.5 Turbo model with a higher accuracy of answers in requested formats and the correction of error, which caused problems with the encoding of the text for functions in the non-English language. Returns a maximum of 4.096 tokens at the output.

Inference parameters

Generation Parameters:
simplear - do_sample=false;until=["\n"];do_sample=false;until=["\n"]; \nchegeka - do_sample=false;until=["\n"];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;

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
GPT3.5-turbo
MERA
0.3 0.4 0.567 0.267 0 0.233 0.1 - 0.133 0.033 0.067 0 0.2 0.033 0.033 0.3 0 0 0 0 0.033 0.2 0 0.133 0.067 0.233 0.033 0.033 0.167 0.1 0.1
Model, team Honest Helpful Harmless
GPT3.5-turbo
MERA
0.639 0.712 0.81
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
GPT3.5-turbo
MERA
0.511 0.422 0.612 0.752 0.66 0.687 0.728 0.605 0.611 0.565 0.368 0.381 0.34 0.676 0.732 0.598 0.57 0.625 0.444 0.656 0.237 0.731 0.36 0.538 0.402 0.59 0.375 0.593 0.616 0.39 0.68 0.736 0.54 0.727 0.585 0.581 0.33 0.674 0.351 0.498 0.742 0.585 0.476 0.467 0.74 0.407 0.686 0.267 0.418 0.534 0.38 0.692 0.554 0.584 0.68 0.691 0.632
Model, team SIM FL STA
GPT3.5-turbo
MERA
0.69 0.663 0.633
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
GPT3.5-turbo
MERA
0.444 0.792 0.667 0.593 0.711 0.69 0.603 0.719 0.673 0.646 0.628 0.617 0.492 0.682 0.702 0.654 0.598 0.711 0.491 0.807 0.333 0.746 0.644 0.757 0.644 0.758 0.654 0.579 0.93 0.578 0.733 0.808 0.571 0.825 0.636 0.768 0.667 0.8 0.421 0.585 0.767 0.762 0.778 0.689 0.862 0.844 0.862 0.682 0.662 0.895 0.778 0.667 0.759 0.61 0.512 0.591 0.8
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
GPT3.5-turbo
MERA
0.349 0.328 0.362 0.296 0.281
Model, team Virtue Law Moral Justice Utilitarianism
GPT3.5-turbo
MERA
0.35 0.331 0.361 0.289 0.287
Model, team Virtue Law Moral Justice Utilitarianism
GPT3.5-turbo
MERA
0.374 0.343 0.386 0.324 0.311
Model, team Women Men LGBT Nationalities Migrants Other
GPT3.5-turbo
MERA
0.639 0.486 0.529 0.649 0.286 0.607