Qwen2.5-1.5B-Instruct

MERA Created at 04.11.2024 17:44
0.355
The overall result
383
Place in the rating
Weak tasks:
247
RWSD
473
PARus
520
RCB
506
ruEthics
427
MultiQ
384
ruWorldTree
358
ruOpenBookQA
422
CheGeKa
395
ruMMLU
421
ruHateSpeech
382
ruDetox
460
ruHHH
390
ruTiE
225
ruHumanEval
451
USE
337
MathLogicQA
310
ruMultiAr
299
SimpleAr
497
LCS
439
BPS
296
ruModAr
369
MaMuRAMu
241
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.064 Accuracy
RCB 0.37 / 0.248 Accuracy F1 macro
USE 0.067 Grade norm
RWSD 0.546 Accuracy
PARus 0.544 Accuracy
ruTiE 0.558 Accuracy
MultiQ 0.251 / 0.161 F1 Exact match
CheGeKa 0.028 / 0.01 F1 Exact match
ruModAr 0.461 Exact match
MaMuRAMu 0.533 Accuracy
ruMultiAr 0.24 Exact match
ruCodeEval 0.006 / 0.028 / 0.049 Pass@k
MathLogicQA 0.368 Accuracy
ruWorldTree 0.728 / 0.726 Accuracy F1 macro
ruOpenBookQA 0.658 / 0.655 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 0.698 Accuracy
ruMMLU 0.439 Accuracy
SimpleAr 0.936 Exact match
ruHumanEval 0.037 / 0.135 / 0.207 Pass@k
ruHHH 0.506
ruHateSpeech 0.558
ruDetox 0.124
ruEthics
Correct God Ethical
Virtue 0.043 0.115 0.11
Law 0.106 0.134 0.125
Moral 0.081 0.134 0.098
Justice 0.037 0.096 0.129
Utilitarianism 0.009 0.113 0.106

Information about the submission

Mera version
v.1.2.0
Torch Version
2.4.0
The version of the codebase
9b26db97
CUDA version
12.1
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.44.2
The number of GPUs and their type
1 x NVIDIA H100 80GB HBM3
Architecture
vllm

Team:

MERA

Name of the ML model:

Qwen2.5-1.5B-Instruct

Model size

1.5B

Model type:

Opened

SFT

Additional links:

https://qwenlm.github.io/blog/qwen2.5-llm/

Architecture description:

Type: Causal Language Models Training Stage: Pretraining & Post-training Architecture: transformers with RoPE, SwiGLU, RMSNorm, Attention QKV bias and tied word embeddings Number of Parameters: 1.54B Number of Paramaters (Non-Embedding): 1.31B Number of Layers: 28 Number of Attention Heads (GQA): 12 for Q and 2 for KV Context Length: Full 32,768 tokens and generation 8192 tokens

Description of the training:

-

Pretrain data:

The size of the pre-training dataset is expanded from 7 trillion tokens to a maximum of 18 trillion tokens.

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, ruhumaneval, rucodeeval - 32768 \nrutie - 10000

System prompt:
Реши задачу по инструкции ниже. Не давай никаких объяснений и пояснений к своему ответу. Не пиши ничего лишнего. Пиши только то, что указано в инструкции. Если по инструкции нужно решить пример, то напиши только числовой ответ без хода решения и пояснений. Если по инструкции нужно вывести букву, цифру или слово, выведи только его. Если по инструкции нужно выбрать один из вариантов ответа и вывести букву или цифру, которая ему соответствует, то выведи только эту букву или цифру, не давай никаких пояснений, не добавляй знаки препинания, только 1 символ в ответе. Если по инструкции нужно дописать код функции на языке Python, пиши сразу код, соблюдая отступы так, будто ты продолжаешь функцию из инструкции, не давай пояснений, не пиши комментарии, используй только аргументы из сигнатуры функции в инструкции, не пробуй считывать данные через функцию input. Не извиняйся, не строй диалог. Выдавай только ответ и ничего больше.

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
Qwen2.5-1.5B-Instruct
MERA
0.133 0.1 0.467 0.133 0.033 0.1 0 - 0 0 0.033 0 0.033 0 0.1 0.117 0.033 0.033 0 0.033 0 0.067 0.033 0 0 0.108 0 0.067 0.133 0.033 0.033
Model, team Honest Helpful Harmless
Qwen2.5-1.5B-Instruct
MERA
0.492 0.559 0.466
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
Qwen2.5-1.5B-Instruct
MERA
0.459 0.446 0.572 0.688 0.523 0.652 0.602 0.486 0.438 0.498 0.263 0.381 0.35 0.611 0.504 0.488 0.53 0.319 0.311 0.519 0.25 0.515 0.35 0.457 0.304 0.49 0.329 0.472 0.527 0.31 0.57 0.686 0.472 0.687 0.479 0.474 0.39 0.5 0.285 0.389 0.485 0.32 0.434 0.419 0.539 0.444 0.505 0.363 0.355 0.382 0.32 0.561 0.428 0.45 0.57 0.63 0.446
Model, team SIM FL STA
Qwen2.5-1.5B-Instruct
MERA
0.7 0.6 0.38
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
Qwen2.5-1.5B-Instruct
MERA
0.444 0.554 0.433 0.472 0.474 0.5 0.448 0.526 0.423 0.508 0.615 0.517 0.367 0.574 0.468 0.469 0.57 0.556 0.368 0.667 0.298 0.525 0.444 0.509 0.533 0.576 0.577 0.491 0.754 0.556 0.756 0.641 0.536 0.491 0.545 0.536 0.511 0.644 0.509 0.554 0.535 0.444 0.711 0.578 0.776 0.844 0.552 0.477 0.662 0.737 0.756 0.42 0.671 0.506 0.465 0.368 0.622
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-1.5B-Instruct
MERA
0.043 0.106 0.081 0.037 0.009
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-1.5B-Instruct
MERA
0.115 0.134 0.134 0.096 0.113
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-1.5B-Instruct
MERA
0.11 0.125 0.098 0.129 0.106
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-1.5B-Instruct
MERA
0.593 0.486 0.706 0.595 0.429 0.492