Qwen2.5-7B-Instruct

MERA Created at 04.11.2024 17:52
0.493
The overall result
158
Place in the rating
Weak tasks:
76
RWSD
223
PARus
110
RCB
263
ruEthics
242
MultiQ
217
ruWorldTree
208
ruOpenBookQA
294
CheGeKa
227
ruMMLU
246
ruHateSpeech
156
ruDetox
117
ruHHH
197
ruTiE
169
ruHumanEval
164
USE
174
MathLogicQA
170
ruMultiAr
207
SimpleAr
305
LCS
293
BPS
153
ruModAr
219
MaMuRAMu
186
ruCodeEval
+19
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.102 Accuracy
RCB 0.566 / 0.493 Accuracy F1 macro
USE 0.208 Grade norm
RWSD 0.619 Accuracy
PARus 0.832 Accuracy
ruTiE 0.711 Accuracy
MultiQ 0.382 / 0.264 F1 Exact match
CheGeKa 0.095 / 0.072 F1 Exact match
ruModAr 0.564 Exact match
MaMuRAMu 0.702 Accuracy
ruMultiAr 0.306 Exact match
ruCodeEval 0.053 / 0.21 / 0.335 Pass@k
MathLogicQA 0.465 Accuracy
ruWorldTree 0.92 / 0.919 Accuracy F1 macro
ruOpenBookQA 0.828 / 0.828 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 0.916 Accuracy
ruMMLU 0.59 Accuracy
SimpleAr 0.974 Exact match
ruHumanEval 0.091 / 0.27 / 0.378 Pass@k
ruHHH 0.815
ruHateSpeech 0.725
ruDetox 0.231
ruEthics
Correct God Ethical
Virtue 0.253 0.284 0.275
Law 0.248 0.261 0.251
Moral 0.267 0.296 0.287
Justice 0.214 0.25 0.24
Utilitarianism 0.224 0.257 0.231

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:

Qwen2.5-7B-Instruct

Model size

7.0B

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, and Attention QKV bias Number of Parameters: 7.61B Number of Paramaters (Non-Embedding): 6.53B Number of Layers: 28 Number of Attention Heads (GQA): 28 for Q and 4 for KV Context Length: Full 131,072 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 - 5000

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
Qwen2.5-7B-Instruct
MERA
0.333 0.367 0.8 0.3 0.1 0.233 0.033 - 0 0.067 0 0.033 0.133 0 0.1 0.217 0.033 0.033 0 0.033 0 0.533 0.233 0.1 0.033 0.392 0.133 0.267 0.367 0.4 0.4
Model, team Honest Helpful Harmless
Qwen2.5-7B-Instruct
MERA
0.787 0.763 0.897
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
Qwen2.5-7B-Instruct
MERA
0.556 0.458 0.796 0.769 0.654 0.731 0.709 0.65 0.636 0.614 0.474 0.508 0.39 0.694 0.708 0.673 0.68 0.611 0.411 0.672 0.251 0.76 0.4 0.555 0.518 0.7 0.406 0.556 0.653 0.42 0.75 0.769 0.632 0.788 0.653 0.662 0.41 0.748 0.543 0.581 0.722 0.577 0.648 0.589 0.747 0.625 0.691 0.493 0.397 0.593 0.58 0.789 0.667 0.693 0.78 0.739 0.663
Model, team SIM FL STA
Qwen2.5-7B-Instruct
MERA
0.495 0.682 0.748
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
Qwen2.5-7B-Instruct
MERA
0.622 0.812 0.6 0.602 0.697 0.672 0.569 0.702 0.692 0.662 0.705 0.725 0.508 0.705 0.602 0.691 0.738 0.667 0.614 0.789 0.649 0.763 0.689 0.74 0.711 0.667 0.718 0.649 0.895 0.689 0.822 0.795 0.634 0.789 0.621 0.696 0.822 0.778 0.579 0.646 0.78 0.762 0.733 0.689 0.879 0.911 0.862 0.795 0.692 0.825 0.8 0.739 0.772 0.61 0.581 0.538 0.789
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-7B-Instruct
MERA
0.253 0.248 0.267 0.214 0.224
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-7B-Instruct
MERA
0.284 0.261 0.296 0.25 0.257
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-7B-Instruct
MERA
0.275 0.251 0.287 0.24 0.231
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-7B-Instruct
MERA
0.769 0.657 0.765 0.676 0.571 0.721