Llama-2-7b-hf

MERA Created at 20.09.2024 15:27
0.297
The overall result
489
Place in the rating
Weak tasks:
316
RWSD
529
PARus
481
RCB
458
MultiQ
531
ruWorldTree
541
ruOpenBookQA
330
CheGeKa
531
ruMMLU
558
ruHateSpeech
550
ruDetox
497
ruHHH
517
ruTiE
162
ruHumanEval
528
USE
525
MathLogicQA
469
ruMultiAr
427
SimpleAr
370
LCS
505
BPS
390
ruModAr
505
MaMuRAMu
150
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.092 Accuracy
RCB 0.409 / 0.402 Accuracy F1 macro
USE 0.048 Grade norm
RWSD 0.523 Accuracy
PARus 0.502 Accuracy
ruTiE 0.504 Accuracy
MultiQ 0.229 / 0.101 F1 Exact match
CheGeKa 0.069 / 0.05 F1 Exact match
ruModAr 0.357 Exact match
MaMuRAMu 0.406 Accuracy
ruMultiAr 0.107 Exact match
ruCodeEval 0.092 / 0.21 / 0.256 Pass@k
MathLogicQA 0.281 Accuracy
ruWorldTree 0.442 / 0.43 Accuracy F1 macro
ruOpenBookQA 0.39 / 0.371 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 0.548 Accuracy
ruMMLU 0.324 Accuracy
SimpleAr 0.792 Exact match
ruHumanEval 0.104 / 0.197 / 0.244 Pass@k
ruHHH 0.489
ruHateSpeech 0.479
ruDetox 0.055
ruEthics
Correct God Ethical
Virtue 0 0.14 -0.028
Law 0 0.076 -0.019
Moral 0 0.132 -0.032
Justice 0 0.122 -0.015
Utilitarianism 0 0.076 -0.005

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-hf

Model size

6.7B

Model type:

Opened

Pretrain

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 generative text model. 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-hf
MERA
0.067 0.033 0.333 0.1 0 0.067 0 - 0 0 0 0 0 0 0.067 0.117 0.033 0.033 0.033 0.033 0 0 0.033 0.033 0.067 0 0.133 0.033 0.1 0.033 0.167
Model, team Honest Helpful Harmless
Llama-2-7b-hf
MERA
0.443 0.525 0.5
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-hf
MERA
0.37 0.331 0.329 0.419 0.399 0.488 0.369 0.354 0.302 0.368 0.246 0.31 0.37 0.315 0.388 0.306 0.36 0.236 0.167 0.351 0.272 0.433 0.32 0.225 0.339 0.3 0.292 0.343 0.343 0.23 0.32 0.512 0.35 0.444 0.332 0.355 0.36 0.316 0.285 0.266 0.333 0.324 0.283 0.255 0.348 0.343 0.343 0.267 0.287 0.307 0.27 0.363 0.308 0.298 0.31 0.339 0.326
Model, team SIM FL STA
Llama-2-7b-hf
MERA
0.879 0.558 0.124
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-hf
MERA
0.333 0.386 0.383 0.389 0.303 0.31 0.431 0.298 0.327 0.492 0.385 0.392 0.333 0.45 0.281 0.469 0.486 0.444 0.316 0.421 0.281 0.542 0.356 0.391 0.4 0.5 0.474 0.351 0.702 0.422 0.578 0.423 0.357 0.368 0.47 0.321 0.356 0.511 0.298 0.385 0.441 0.254 0.533 0.244 0.586 0.378 0.5 0.273 0.446 0.491 0.4 0.435 0.532 0.429 0.326 0.368 0.478
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-hf
MERA
0 0 0 0 0
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-hf
MERA
0.14 0.076 0.132 0.122 0.076
Model, team Virtue Law Moral Justice Utilitarianism
Llama-2-7b-hf
MERA
-0.028 -0.019 -0.032 -0.015 -0.005
Model, team Women Men LGBT Nationalities Migrants Other
Llama-2-7b-hf
MERA
0.481 0.314 0.588 0.432 0.714 0.541