Qwen2.5-0.5B-Instruct

MERA Created at 04.11.2024 17:43
0.286
The overall result
502
Place in the rating
Weak tasks:
496
RWSD
592
PARus
551
RCB
525
MultiQ
527
ruWorldTree
531
ruOpenBookQA
596
CheGeKa
513
ruMMLU
484
ruHateSpeech
607
ruDetox
552
ruHHH
496
ruTiE
158
ruHumanEval
567
USE
464
MathLogicQA
380
ruMultiAr
393
SimpleAr
516
LCS
544
BPS
362
ruModAr
523
MaMuRAMu
168
ruCodeEval
+18
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.056 Accuracy
RCB 0.347 / 0.277 Accuracy F1 macro
USE 0.037 Grade norm
RWSD 0.477 Accuracy
PARus 0.47 Accuracy
ruTiE 0.507 Accuracy
MultiQ 0.161 / 0.089 F1 Exact match
CheGeKa 0.004 / 0 F1 Exact match
ruModAr 0.397 Exact match
MaMuRAMu 0.383 Accuracy
ruMultiAr 0.186 Exact match
ruCodeEval 0.076 / 0.212 / 0.287 Pass@k
MathLogicQA 0.312 Accuracy
ruWorldTree 0.453 / 0.437 Accuracy F1 macro
ruOpenBookQA 0.41 / 0.382 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 0.51 Accuracy
ruMMLU 0.341 Accuracy
SimpleAr 0.854 Exact match
ruHumanEval 0.108 / 0.242 / 0.287 Pass@k
ruHHH 0.466
ruHateSpeech 0.528
ruDetox 0.042
ruEthics
Correct God Ethical
Virtue -0.008 -0.057 -0.041
Law -0.019 -0.036 -0.044
Moral -0.012 -0.039 -0.038
Justice -0 -0.041 -0.058
Utilitarianism 0.017 -0.028 -0.033

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:

Qwen2.5-0.5B-Instruct

Model size

500.0M

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: 0.49B Number of Paramaters (Non-Embedding): 0.36B Number of Layers: 24 Number of Attention Heads (GQA): 14 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:
chegeka - 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; \nsimplear - do_sample=false;until=["\n"];

The size of the context:
bps, lcs, chegeka, mathlogicqa, parus, rcb, rudetox, ruhatespeech, rummlu, ruworldtree, ruopenbookqa, rumultiar, use, rwsd, mamuramu, multiq, rumodar, ruethics, ruhhh, ruhumaneval, rucodeeval, simplear - 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
Qwen2.5-0.5B-Instruct
MERA
0.133 0 0.133 0.1 0 0.033 0 - 0 0 0 0 0 0 0 0.283 0.067 0 0 0.033 0 0 0 0 0 0.008 0.033 0.033 0.067 0 0.033
Model, team Honest Helpful Harmless
Qwen2.5-0.5B-Instruct
MERA
0.426 0.492 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
Qwen2.5-0.5B-Instruct
MERA
0.393 0.319 0.388 0.513 0.379 0.373 0.417 0.447 0.352 0.332 0.184 0.262 0.33 0.454 0.383 0.396 0.44 0.229 0.233 0.389 0.249 0.327 0.3 0.393 0.268 0.34 0.309 0.389 0.327 0.27 0.41 0.545 0.362 0.545 0.366 0.316 0.29 0.335 0.285 0.379 0.364 0.228 0.428 0.321 0.358 0.296 0.417 0.27 0.305 0.328 0.3 0.401 0.269 0.315 0.46 0.412 0.332
Model, team SIM FL STA
Qwen2.5-0.5B-Instruct
MERA
0.811 0.549 0.134
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
Qwen2.5-0.5B-Instruct
MERA
0.311 0.416 0.383 0.417 0.355 0.276 0.241 0.228 0.385 0.431 0.449 0.408 0.258 0.45 0.275 0.37 0.533 0.356 0.404 0.404 0.211 0.305 0.289 0.402 0.467 0.318 0.41 0.263 0.614 0.244 0.622 0.436 0.402 0.421 0.333 0.357 0.333 0.578 0.351 0.323 0.318 0.349 0.422 0.444 0.655 0.644 0.328 0.432 0.4 0.544 0.578 0.319 0.443 0.39 0.372 0.292 0.367
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
-0.008 -0.019 -0.012 -0 0.017
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
-0.057 -0.036 -0.039 -0.041 -0.028
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
-0.041 -0.044 -0.038 -0.058 -0.033
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-0.5B-Instruct
MERA
0.5 0.714 0.529 0.595 0.286 0.459