Qwen2.5-0.5B-Instruct

MERA Created at 04.11.2024 17:41
0.259
The overall result
523
Place in the rating
Weak tasks:
492
RWSD
597
PARus
592
RCB
529
MultiQ
512
ruWorldTree
502
ruOpenBookQA
486
CheGeKa
504
ruMMLU
552
ruHateSpeech
505
ruDetox
593
ruHHH
428
ruTiE
440
USE
462
MathLogicQA
378
ruMultiAr
389
SimpleAr
460
LCS
491
BPS
503
ruModAr
513
MaMuRAMu
+16
Hide

Ratings for leaderboard tasks

The table will scroll to the left

Task name Result Metric
LCS 0.074 Accuracy
RCB 0.329 / 0.246 Accuracy F1 macro
USE 0.068 Grade norm
RWSD 0.477 Accuracy
PARus 0.458 Accuracy
ruTiE 0.529 Accuracy
MultiQ 0.146 / 0.097 F1 Exact match
CheGeKa 0.015 / 0 F1 Exact match
ruModAr 0.064 Exact match
MaMuRAMu 0.386 Accuracy
ruMultiAr 0.182 Exact match
ruCodeEval 0 / 0 / 0 Pass@k
MathLogicQA 0.31 Accuracy
ruWorldTree 0.474 / 0.468 Accuracy F1 macro
ruOpenBookQA 0.463 / 0.441 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 0.553 Accuracy
ruMMLU 0.342 Accuracy
SimpleAr 0.851 Exact match
ruHumanEval 0 / 0 / 0 Pass@k
ruHHH 0.348
ruHateSpeech 0.475
ruDetox 0.068
ruEthics
Correct God Ethical
Virtue 0.015 -0.007 -0.028
Law -0.014 -0.045 -0.066
Moral -0 -0.023 -0.028
Justice 0.011 -0.02 -0.05
Utilitarianism 0.01 0.033 -0.028

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:
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 - 15000

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.

Description of the template:
{%- if tools %} \n {{- '<|im_start|>system\n' }} \n {%- if messages[0]['role'] == 'system' %} \n {{- messages[0]['content'] }} \n {%- else %} \n {{- 'You are Qwen, created by Alibaba Cloud. You are a helpful assistant.' }} \n {%- endif %} \n {{- "\n\n# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }} \n {%- for tool in tools %} \n {{- "\n" }} \n {{- tool | tojson }} \n {%- endfor %} \n {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }} \n{%- else %} \n {%- if messages[0]['role'] == 'system' %} \n {{- '<|im_start|>system\n' + messages[0]['content'] + '<|im_end|>\n' }} \n {%- else %} \n {{- '<|im_start|>system\nYou are Qwen, created by Alibaba Cloud. You are a helpful assistant.<|im_end|>\n' }} \n {%- endif %} \n{%- endif %} \n{%- for message in messages %} \n {%- if (message.role == "user") or (message.role == "system" and not loop.first) or (message.role == "assistant" and not message.tool_calls) %} \n {{- '<|im_start|>' + message.role + '\n' + message.content + '<|im_end|>' + '\n' }} \n {%- elif message.role == "assistant" %} \n {{- '<|im_start|>' + message.role }} \n {%- if message.content %} \n {{- '\n' + message.content }} \n {%- endif %} \n {%- for tool_call in message.tool_calls %} \n {%- if tool_call.function is defined %} \n {%- set tool_call = tool_call.function %} \n {%- endif %} \n {{- '\n<tool_call>\n{"name": "' }} \n {{- tool_call.name }} \n {{- '", "arguments": ' }} \n {{- tool_call.arguments | tojson }} \n {{- '}\n</tool_call>' }} \n {%- endfor %} \n {{- '<|im_end|>\n' }} \n {%- elif message.role == "tool" %} \n {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != "tool") %} \n {{- '<|im_start|>user' }} \n {%- endif %} \n {{- '\n<tool_response>\n' }} \n {{- message.content }} \n {{- '\n</tool_response>' }} \n {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %} \n {{- '<|im_end|>\n' }} \n {%- endif %} \n {%- endif %} \n{%- endfor %} \n{%- if add_generation_prompt %} \n {{- '<|im_start|>assistant\n' }} \n{%- endif %}

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
0.167 0 0.133 0.1 0 0.033 0 - 0.033 0.033 0 0.033 0 0 0 0.483 0.067 0 0.033 0.033 0.033 0.033 0 0 0 0.05 0.133 0.133 0.067 0 0.067
Model, team Honest Helpful Harmless
Qwen2.5-0.5B-Instruct
MERA
0.361 0.373 0.31
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
0.407 0.361 0.368 0.556 0.408 0.388 0.388 0.428 0.33 0.359 0.246 0.262 0.34 0.38 0.359 0.396 0.46 0.236 0.222 0.389 0.247 0.251 0.3 0.347 0.286 0.38 0.304 0.426 0.322 0.24 0.45 0.529 0.387 0.515 0.366 0.393 0.27 0.352 0.272 0.315 0.409 0.217 0.448 0.324 0.378 0.329 0.373 0.304 0.309 0.328 0.31 0.401 0.285 0.328 0.4 0.43 0.326
Model, team SIM FL STA
Qwen2.5-0.5B-Instruct
MERA
0.38 0.611 0.527
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 0.475 0.4 0.407 0.408 0.379 0.328 0.246 0.308 0.415 0.449 0.383 0.258 0.442 0.275 0.383 0.439 0.356 0.316 0.421 0.228 0.356 0.289 0.379 0.444 0.288 0.397 0.298 0.579 0.333 0.689 0.449 0.393 0.333 0.409 0.446 0.378 0.578 0.351 0.323 0.31 0.302 0.444 0.467 0.638 0.711 0.414 0.477 0.492 0.491 0.533 0.319 0.481 0.403 0.419 0.234 0.367
Coorect
Good
Ethical
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
0.015 -0.014 -0 0.011 0.01
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
-0.007 -0.045 -0.023 -0.02 0.033
Model, team Virtue Law Moral Justice Utilitarianism
Qwen2.5-0.5B-Instruct
MERA
-0.028 -0.066 -0.028 -0.05 -0.028
Model, team Women Men LGBT Nationalities Migrants Other
Qwen2.5-0.5B-Instruct
MERA
0.472 0.543 0.529 0.568 0.143 0.41