Llama-2-7b-chat-hf

MERA Created at 22.09.2024 20:43
0.3
The overall result
486
Place in the rating
Weak tasks:
133
RWSD
500
PARus
483
RCB
371
ruEthics
432
MultiQ
509
ruWorldTree
514
ruOpenBookQA
387
CheGeKa
517
ruMMLU
548
ruHateSpeech
384
ruDetox
505
ruHHH
501
ruTiE
152
ruHumanEval
454
USE
529
MathLogicQA
471
ruMultiAr
443
SimpleAr
383
LCS
567
BPS
410
ruModAr
515
MaMuRAMu
190
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.09 Accuracy
RCB 0.406 / 0.361 Accuracy F1 macro
USE 0.067 Grade norm
RWSD 0.585 Accuracy
PARus 0.52 Accuracy
ruTiE 0.506 Accuracy
MultiQ 0.245 / 0.129 F1 Exact match
CheGeKa 0.036 / 0.017 F1 Exact match
ruModAr 0.31 Exact match
MaMuRAMu 0.395 Accuracy
ruMultiAr 0.104 Exact match
ruCodeEval 0.048 / 0.089 / 0.104 Pass@k
MathLogicQA 0.279 Accuracy
ruWorldTree 0.526 / 0.513 Accuracy F1 macro
ruOpenBookQA 0.453 / 0.432 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 0.491 Accuracy
ruMMLU 0.334 Accuracy
SimpleAr 0.702 Exact match
ruHumanEval 0.114 / 0.176 / 0.195 Pass@k
ruHHH 0.483
ruHateSpeech 0.487
ruDetox 0.124
ruEthics
Correct God Ethical
Virtue 0.15 0.171 0.221
Law 0.17 0.235 0.269
Moral 0.14 0.192 0.276
Justice 0.126 0.23 0.225
Utilitarianism 0.154 0.196 0.195

Information about the submission

Mera version
v.1.2.0
Torch Version
2.3.1
The version of the codebase
aec92f8
CUDA version
12.1
Precision of the model weights
float16
Seed
1234
Batch
1
Transformers version
4.42.3
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
hf

Team:

MERA

Name of the ML model:

Llama-2-7b-chat-hf

Model size

6.7B

Model type:

Opened

SFT

Additional links:

https://arxiv.org/pdf/2307.09288

Architecture description:

Auto-regressive language model that uses an optimized transformer architecture, with context length of 4K tokens.

Description of the training:

Pretrained and fine-tuned (SFT & RLHF) generative text model, optimized for dialogue use cases. The global batch-size is 4M tokens. The model was trained between January 2023 and July 2023.

Pretrain data:

A new mix of publicly available online data, the pretraining data volume is 2.0T.

License:

https://ai.meta.com/resources/models-and-libraries/llama-downloads/

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
Llama-2-7b-chat-hf
MERA
0.033 0.067 0.467 0.1 0 0.067 0 - 0 0 0 0.033 0.033 0 0.067 0.167 0.067 0.033 0.033 0.033 0.033 0 0 0 0.033 0.05 0.067 0.133 0.167 0.133 0.133
Model, team Honest Helpful Harmless
Llama-2-7b-chat-hf
MERA
0.475 0.542 0.431
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
Llama-2-7b-chat-hf
MERA
0.407 0.41 0.349 0.504 0.34 0.383 0.417 0.399 0.37 0.426 0.263 0.222 0.34 0.454 0.464 0.358 0.35 0.278 0.189 0.336 0.24 0.333 0.29 0.237 0.295 0.34 0.272 0.417 0.269 0.25 0.42 0.529 0.313 0.495 0.351 0.321 0.27 0.365 0.272 0.315 0.404 0.272 0.345 0.265 0.375 0.208 0.387 0.289 0.28 0.339 0.2 0.439 0.292 0.298 0.27 0.467 0.383
Model, team SIM FL STA
Llama-2-7b-chat-hf
MERA
0.481 0.557 0.596
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
Llama-2-7b-chat-hf
MERA
0.289 0.416 0.383 0.296 0.368 0.397 0.328 0.333 0.462 0.462 0.372 0.333 0.367 0.372 0.357 0.309 0.467 0.356 0.281 0.386 0.298 0.424 0.378 0.45 0.333 0.424 0.385 0.351 0.737 0.333 0.622 0.449 0.393 0.439 0.364 0.339 0.289 0.533 0.316 0.338 0.453 0.286 0.467 0.289 0.586 0.489 0.431 0.295 0.431 0.456 0.4 0.449 0.506 0.416 0.256 0.316 0.456
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-chat-hf
MERA
0.15 0.17 0.14 0.126 0.154
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-chat-hf
MERA
0.171 0.235 0.192 0.23 0.196
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-chat-hf
MERA
0.221 0.269 0.276 0.225 0.195
Model, team Women Men LGBT Nationalities Migrants Other
Llama-2-7b-chat-hf
MERA
0.463 0.457 0.588 0.432 0.429 0.557