evalscope 0.10.0__py3-none-any.whl → 1.2.0__py3-none-any.whl

evalscope/models/model.py

@@ -1,229 +0,0 @@
-# Copyright (c) Alibaba, Inc. and its affiliates.
-import os
-import random
-import time
-from abc import ABC, abstractmethod
-from typing import Any
-
-from evalscope.utils.logger import get_logger
-
-logger = get_logger()
-
-
-class BaseModel(ABC):
-
-    def __init__(self, model_cfg: dict, **kwargs):
-        """
-        Base model class.
-
-        Args:
-            model_cfg (dict): The model configuration. Depending on the specific model. Example:
-                {'model_id': 'modelscope/Llama-2-7b-chat-ms', 'revision': 'v1.0.0'}
-
-            **kwargs: kwargs
-        """
-        self.model_cfg: dict = model_cfg
-        self.kwargs = kwargs
-
-    @abstractmethod
-    def predict(self, *args, **kwargs) -> Any:
-        """
-        Model prediction func.
-        """
-        raise NotImplementedError
-
-
-class ChatBaseModel(BaseModel):
-
-    def __init__(self, model_cfg: dict, **kwargs):
-        """
-        Chat base model class. Depending on the specific model.
-
-        Args:
-            model_cfg (dict):
-                {'model_id': 'modelscope/Llama-2-7b-chat-ms', 'revision': 'v1.0.0', 'device_map': 'auto'}
-
-            **kwargs: kwargs
-        """
-        super(ChatBaseModel, self).__init__(model_cfg=model_cfg, **kwargs)
-
-    @abstractmethod
-    def predict(self, inputs: dict, **kwargs) -> dict:
-        """
-        Model prediction func. The inputs and outputs are compatible with OpenAI Chat Completions APIs.
-        Refer to: https://platform.openai.com/docs/guides/gpt/chat-completions-api
-
-        # TODO: follow latest OpenAI API
-
-        Args:
-            inputs (dict): The input prompts and history. Input format:
-                {'messages': [
-                    {'role': 'system', 'content': 'You are a helpful assistant.'},
-                    {'role': 'user', 'content': 'Who won the world series in 2020?'},
-                    {'role': 'assistant', 'content': 'The Los Angeles Dodgers won the World Series in 2020.'},
-                ]
-                'history': [
-                    {'role': 'system', 'content': 'Hello'},
-                    {'role': 'user', 'content': 'Hi'}]
-                }
-
-            kwargs (dict): Could be inference configuration. Default: None.
-                cfg format: {'max_length': 1024}
-
-        Returns: The result format:
-                {
-                  'choices': [
-                    {
-                      'index': 0,
-                      'message': {
-                        'content': 'The 2020 World Series was played in Texas at Globe Life Field in Arlington.',
-                        'role': 'assistant'
-                      }
-                    }
-                  ],
-                  'created': 1677664795,
-                  # For models on the ModelScope or HuggingFace, concat model_id and revision with "-".
-                  'model': 'gpt-3.5-turbo-0613',
-                  'object': 'chat.completion',
-                  'usage': {
-                    'completion_tokens': 17,
-                    'prompt_tokens': 57,
-                    'total_tokens': 74
-                  }
-                }
-        """
-        raise NotImplementedError
-
-
-class OpenAIModel(ChatBaseModel):
-    """
-    APIs of OpenAI models.
-    Available models: gpt-3.5-turbo, gpt-4
-    """
-
-    MAX_RETRIES = 3
-
-    def __init__(self, model_cfg: dict, **kwargs):
-        super(OpenAIModel, self).__init__(model_cfg=model_cfg, **kwargs)
-
-        openai_api_key = os.environ.get('OPENAI_API_KEY', None)
-        self.api_key = self.model_cfg.get('api_key', openai_api_key)
-
-        if not self.api_key:
-            logger.error('OpenAI API key is not provided, please set it in environment variable OPENAI_API_KEY')
-            # raise ValueError(
-            #     'OpenAI API key is not provided, '
-            #     'please set it in environment variable OPENAI_API_KEY')
-
-    def predict(self, model_id: str, inputs: dict, **kwargs) -> dict:
-
-        sys_prompt: str = inputs.get('sys_prompt', '')
-        user_prompt: str = inputs.get('user_prompt', '')
-
-        # model_id: str = kwargs.get('model_id', '')
-        temperature: float = kwargs.pop('temperature', 0.2)
-        max_tokens: int = kwargs.pop('max_tokens', 1024)
-        mode: str = kwargs.pop('mode', 'chat.completion')
-
-        logger.info(f'Using OpenAI model_id: {model_id}')
-
-        res = self._predict(
-            model_id=model_id,
-            sys_prompt=sys_prompt,
-            user_prompt=user_prompt,
-            temperature=temperature,
-            max_tokens=max_tokens,
-            mode=mode)
-
-        return res
-
-    def _predict(
-        self,
-        model_id,
-        sys_prompt,
-        user_prompt,
-        temperature,
-        max_tokens,
-        mode: str = 'chat.completion',
-    ) -> dict:
-        import openai
-
-        res = {}
-        openai.api_key = self.api_key
-
-        for i in range(self.MAX_RETRIES):
-            try:
-                if mode == 'chat.completion':
-                    resp = openai.ChatCompletion.create(
-                        model=model_id,
-                        messages=[{
-                            'role': 'system',
-                            'content': sys_prompt
-                        }, {
-                            'role': 'user',
-                            'content': user_prompt
-                        }],
-                        temperature=temperature,
-                        max_tokens=max_tokens)
-
-                    if resp:
-                        ans_text = resp['choices'][0]['message']['content']
-                        model_id = resp['model']
-                    else:
-                        logger.warning(f'OpenAI GPT API call failed: got empty response '
-                                       f'for input {sys_prompt} {user_prompt}')
-                        ans_text = ''
-                        model_id = ''
-
-                    res['ans_text'] = ans_text
-                    res['model_id'] = model_id
-                else:
-                    raise ValueError(f'Invalid mode: {mode}')
-
-                return res
-
-            except Exception as e:
-                logger.warning(f'OpenAI API call failed: {e}')
-                time.sleep(3)
-        logger.error(f'OpenAI API call failed after {self.MAX_RETRIES} retries')
-        return res
-
-
-class DummyChatModel(ChatBaseModel):
-
-    MODEL_ID = 'dummy_chat_model_0801'
-    REVISION = 'v1.0.0'
-
-    def __init__(self, model_cfg: dict, **kwargs):
-        model_cfg['model_id'] = self.MODEL_ID
-        model_cfg['revision'] = self.REVISION
-        super(DummyChatModel, self).__init__(model_cfg=model_cfg)
-
-    def predict(self, inputs: dict, **kwargs) -> dict:
-
-        debug: bool = False
-        if debug:
-            messages = inputs['messages']
-            history = inputs['history']
-
-            logger.info(f'** messages: {messages}')
-            logger.info(f'** history: {history}')
-
-        choice = random.choice(['A', 'B', 'C', 'D'])
-
-        # Build response
-        res = {
-            'choices': [{
-                'index': 0,
-                'message': {
-                    'content': choice,
-                    'role': 'assistant'
-                }
-            }],
-            'created': time.time(),
-            'model': self.MODEL_ID + '-' + self.REVISION,
-            'object': 'chat.completion',
-            'usage': {}
-        }
-
-        return res

evalscope/models/server_adapter.py

@@ -1,111 +0,0 @@
-import requests
-import time
-from typing import Optional, Union
-
-from evalscope.models.base_adapter import BaseModelAdapter
-from evalscope.utils.chat_service import ChatMessage
-from evalscope.utils.logger import get_logger
-
-logger = get_logger()
-
-
-class ServerModelAdapter(BaseModelAdapter):
-    """
-    Server model adapter to request remote API model and generate results.
-    """
-
-    def __init__(self, api_url: str, model_id: str, api_key: str = 'EMPTY', **kwargs):
-        """
-        Args:
-            api_url: The URL of the remote API model.
-            model_id: The ID of the remote API model.
-            api_key: The API key of the remote API model.
-        """
-        self.api_url = api_url
-        self.model_id = model_id
-        self.api_key = api_key
-        self.seed = kwargs.get('seed', None)
-        self.model_cfg = {'api_url': api_url, 'model_id': model_id, 'api_key': api_key}
-        super().__init__(model=None, model_cfg=self.model_cfg, **kwargs)
-
-    def predict(self, inputs: Union[str, dict, list], infer_cfg: dict = None) -> dict:
-        """
-        Model prediction func.
-
-        Args:
-            inputs (Union[str, dict, list]): The input data.
-            infer_cfg (dict): Inference configuration.
-
-        Returns:
-            res (dict): The model prediction results.
-        """
-        infer_cfg = infer_cfg or {}
-
-        # Process inputs
-        if isinstance(inputs, str):
-            query = inputs
-            system_prompt = None
-        elif isinstance(inputs, dict):
-            data: list = inputs['data']
-            if isinstance(data[0], tuple):  # for truthful_qa and hellaswag
-                query = '\n'.join(''.join(item) for item in data)
-                system_prompt = inputs.get('system_prompt', None)
-            else:
-                query = data[0]
-                system_prompt = inputs.get('system_prompt', None)
-        elif isinstance(inputs, list):
-            query = '\n'.join(inputs)
-            system_prompt = None
-        else:
-            raise TypeError(f'Unsupported inputs type: {type(inputs)}')
-
-        content = self.make_request_content(query, system_prompt)
-        request_json = self.make_request(content, infer_cfg)
-        response = self.send_request(request_json)
-        return response
-
-    def make_request_content(self, query: str, system_prompt: Optional[str] = None) -> dict:
-        """
-        Make request content for API.
-        """
-        if system_prompt is not None:
-            messages = [
-                ChatMessage(role='system', content=system_prompt).model_dump(exclude_unset=True),
-                ChatMessage(role='user', content=query).model_dump(exclude_unset=True)
-            ]
-        else:
-            messages = [ChatMessage(role='user', content=query).model_dump(exclude_unset=True)]
-        return {'messages': messages}
-
-    def make_request(self, content: dict, infer_cfg: dict = {}) -> dict:
-        """Make request to remote API."""
-        # Format request JSON according to OpenAI API format
-        do_sample = infer_cfg.get('do_sample', False)
-        temperature = infer_cfg.get('temperature', 0.0) if do_sample else 0.0
-
-        request_json = {
-            **content, 'model': self.model_id,
-            'max_tokens': infer_cfg.get('max_tokens', 2048),
-            'temperature': temperature,
-            'top_p': infer_cfg.get('top_p', 1.0),
-            'n': infer_cfg.get('num_return_sequences', 1),
-            'stop': infer_cfg.get('stop', None)
-        }
-        if self.seed is not None:
-            request_json['seed'] = self.seed
-        logger.debug(f'Request to remote API: {request_json}')
-        return request_json
-
-    def send_request(self, request_json: dict, max_retries: int = 3) -> dict:
-        for attempt in range(max_retries):
-            response = requests.post(
-                self.api_url, json=request_json, headers={'Authorization': f'Bearer {self.api_key}'})
-            if response.status_code == 200:
-                response_data = response.json()
-                return response_data
-            logger.warning(f'Failed to request to remote API: {response.status_code} {response.text}')
-            if attempt < max_retries - 1:
-                time.sleep(5)  # Sleep for 5 seconds before retrying
-            else:
-                raise RuntimeError(f'Failed to request to remote API after {max_retries} attempts: '
-                                   f'{response.status_code} {response.text}')

evalscope/registry/__init__.py

@@ -1 +0,0 @@
-# Copyright (c) Alibaba, Inc. and its affiliates.

evalscope/registry/config/cfg_arena.yaml

@@ -1,77 +0,0 @@
-# input raw data
-question_file: registry/data/question.jsonl
-
-# candidate models to be battled
-answers_gen:
-    chatglm3-6b:
-        # model_id_or_path could be local absolute path, e.g. /to/path/.cache/modelscope/ZhipuAI/chatglm3-6b
-        model_id_or_path: ZhipuAI/chatglm3-6b       # model_id on modelscope
-        revision: v1.0.2        # revision of model, default is NULL
-        precision: torch.float16
-        enable: true            # enable or disable this model
-        template_type: chatglm3  # see: https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E6%94%AF%E6%8C%81%E7%9A%84%E6%A8%A1%E5%9E%8B%E5%92%8C%E6%95%B0%E6%8D%AE%E9%9B%86.md
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.333
-        # output predicted answer file name
-        output_file: registry/data/arena/answers/answer_chatglm3-6b.jsonl
-    Baichuan2-7B-Base:
-        model_id_or_path: baichuan-inc/Baichuan2-7B-Base
-        revision: v1.0.2       # revision of model, default is NULL
-        precision: torch.float16
-        enable: false           # enable or disable this model
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Baichuan2-7B-Base.jsonl
-    Qwen-7B:
-        model_id_or_path: qwen/Qwen-7B
-        revision: v1.1.8       # revision of model, default is NULL
-        precision: torch.float16
-        enable: true           # enable or disable this model      # TODO: tokenizer issue
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Qwen-7B.jsonl
-
-# Auto-reviewer(GPT-4) config
-reviews_gen:
-    enable: true
-    reviewer:
-        # class reference of auto reviewer(GPT-4)
-        ref: evalscope.evaluator.reviewer.auto_reviewer:AutoReviewerGpt4
-        args:
-            max_tokens: 1024
-            temperature: 0.2
-            # options: pairwise, pairwise_baseline, single (default is pairwise)
-            mode: pairwise
-            # position bias mitigation strategy, options: swap_position, randomize_order, NULL. default is NULL
-            position_bias_mitigation: NULL
-            # completion parser config, default is lmsys_parser
-            fn_completion_parser: lmsys_parser
-    # prompt templates for auto reviewer(GPT-4)
-    prompt_file: registry/data/prompt_template/prompt_templates.jsonl
-    # target answer files list to be reviewed,
-    # could be replaced by your own path: ['/path/to/answers_model_1.jsonl', '/path/to/answers_model_2.jsonl', ...]
-    # Default is NULL, which means all answers in answers_gen will be reviewed
-    target_answers: NULL
-    # output file name of auto reviewer
-    review_file: registry/data/arena/reviews/review_gpt4.jsonl
-
-# rating results
-rating_gen:
-    enable: true
-    metrics: ['elo']
-    # elo rating report file name
-    report_file: registry/data/arena/reports/elo_rating_origin.csv

evalscope/registry/config/cfg_arena_zhihu.yaml

@@ -1,63 +0,0 @@
-# input raw data
-question_file: registry/data/question.jsonl
-
-# candidate models to be battled
-answers_gen:
-    Qwen2-7B-Instruct:
-        model_id_or_path: /mnt/data/data/user/maoyunlin.myl/models/Qwen2-7B-Instruct       # model_id on modelscope
-        revision: NULL        # revision of model, default is NULL
-        precision: torch.float16
-        enable: true            # enable or disable this model
-        template_type: default-generation  # see: https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E6%94%AF%E6%8C%81%E7%9A%84%E6%A8%A1%E5%9E%8B%E5%92%8C%E6%95%B0%E6%8D%AE%E9%9B%86.md
-        generation_config:
-            do_sample: true
-            max_new_tokens: 512
-            top_k: 20
-            top_p: 0.9
-            temperature: 0.7
-        # output predicted answer file name
-        output_file: registry/data/arena/answers/answer_qwen2.jsonl
-    Qwen-7B:
-        model_id_or_path: /mnt/data/data/user/maoyunlin.myl/output/qwen2-7b-instruct/v25-20240809-113533/checkpoint-309-merged
-        revision: NULL       # revision of model, default is NULL
-        precision: torch.float16
-        enable: true           # enable or disable this model
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 512
-            top_k: 20
-            top_p: 0.9
-            temperature: 0.7
-        output_file: registry/data/arena/answers/answer_Qwen-7B.jsonl
-
-# Auto-reviewer(GPT-4) config
-reviews_gen:
-    enable: true
-    reviewer:
-        # class reference of auto reviewer(GPT-4)
-        ref: evalscope.evaluator.reviewer.auto_reviewer:AutoReviewerGpt4
-        args:
-            max_tokens: 1024
-            temperature: 0.2
-            # options: pairwise, pairwise_baseline, single (default is pairwise)
-            mode: pairwise
-            # position bias mitigation strategy, options: swap_position, randomize_order, NULL. default is NULL
-            position_bias_mitigation: NULL
-            # completion parser config, default is lmsys_parser
-            fn_completion_parser: lmsys_parser
-    # prompt templates for auto reviewer(GPT-4)
-    prompt_file: registry/data/prompt_template/prompt_templates.jsonl
-    # target answer files list to be reviewed,
-    # could be replaced by your own path: ['/path/to/answers_model_1.jsonl', '/path/to/answers_model_2.jsonl', ...]
-    # Default is NULL, which means all answers in answers_gen will be reviewed
-    target_answers: NULL
-    # output file name of auto reviewer
-    review_file: registry/data/arena/reviews/review_gpt4.jsonl
-
-# rating results
-rating_gen:
-    enable: true
-    metrics: ['elo']
-    # elo rating report file name
-    report_file: registry/data/arena/reports/elo_rating_origin.csv

evalscope/registry/config/cfg_pairwise_baseline.yaml

@@ -1,83 +0,0 @@
-# input raw data
-question_file: registry/data/question.jsonl
-
-# candidate models to be battled
-answers_gen:
-    chatglm3-6b:
-        # model_id_or_path could be local absolute path, e.g. /to/path/.cache/modelscope/ZhipuAI/chatglm3-6b
-        model_id_or_path: ZhipuAI/chatglm3-6b       # model_id on modelscope
-        revision: v1.0.2        # revision of model, default is NULL
-        precision: torch.float16
-        enable: true            # enable or disable this model
-        template_type: chatglm3  # see: https://github.com/modelscope/swift/blob/main/docs/source/LLM/%E6%94%AF%E6%8C%81%E7%9A%84%E6%A8%A1%E5%9E%8B%E5%92%8C%E6%95%B0%E6%8D%AE%E9%9B%86.md
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        # output predicted answer file name
-        output_file: registry/data/arena/answers/answer_chatglm3-6b.jsonl
-    Baichuan2-7B-Base:
-        model_id_or_path: baichuan-inc/Baichuan2-7B-Base
-        revision: v1.0.2       # revision of model, default is NULL
-        precision: torch.float16
-        enable: false           # enable or disable this model
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Baichuan2-7B-Base.jsonl
-    Qwen-7B:
-        model_id_or_path: qwen/Qwen-7B
-        revision: v1.1.8       # revision of model, default is NULL
-        precision: torch.float16
-        enable: true           # enable or disable this model      # TODO: tokenizer issue
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Qwen-7B.jsonl
-
-# model of auto-reviewer
-reviews_gen:
-    enable: true
-    reviewer:
-        ref: evalscope.evaluator.reviewer.auto_reviewer:AutoReviewerGpt4
-        args:
-            model: gpt-4
-            max_tokens: 1024
-            temperature: 0
-            # pairwise comparison against baseline
-            mode: pairwise_baseline
-            # position bias mitigation strategy, options: swap_position, randomize_order, None. default is None
-            position_bias_mitigation: swap_position
-            # completion parser config, default is lmsys_parser
-            fn_completion_parser: lmsys_parser
-    # target answers list to be reviewed, could be replaced by your own path: /path/to/answers.jsonl
-    target_answers: [registry/data/arena/answers/answer_chatglm3-6b.jsonl,
-                     registry/data/arena/answers/answer_Baichuan2-7B-Base.jsonl]
-    # the path to the outputs of the baseline model
-    baseline_file: registry/data/arena/answers/answer_text_davinci_003.jsonl
-    # the path to the reference answers
-    reference_file:
-    # prompt templates for auto reviewer(GPT-4)
-    prompt_file: registry/data/prompt_template/lmsys_v2.jsonl
-    # output file of auto reviewer
-    review_file: registry/data/arena/reviews/review_gpt4_pair_baseline.jsonl
-    # cache file of auto reviewer
-    cache_file: registry/data/arena/reviews/review_gpt4_pair_baseline.jsonl
-
-# rating results
-rating_gen:
-    enable: true
-    metrics: ['pairwise']
-    baseline_model: text_davinci_003
-    # elo rating report file
-    report_file: registry/data/arena/reports/rating_pairwise_baseline.csv

evalscope/registry/config/cfg_single.yaml

@@ -1,78 +0,0 @@
-# input raw data
-question_file: registry/data/question.jsonl
-
-# candidate models to be battled
-answers_gen:
-    chatglm3-6b:
-        # model_id_or_path could be local absolute path, e.g. /to/path/.cache/modelscope/ZhipuAI/chatglm3-6b
-        model_id_or_path: ZhipuAI/chatglm3-6b       # model_id on modelscope
-        revision: v1.0.2        # revision of model, default is NULL
-        precision: torch.float16
-        enable: true            # enable or disable this model
-        template_type: chatglm3
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        # output predicted answer file name
-        output_file: registry/data/arena/answers/answer_chatglm3-6b.jsonl
-    Baichuan2-7B-Base:
-        model_id_or_path: baichuan-inc/Baichuan2-7B-Base
-        revision: v1.0.2       # revision of model, default is NULL
-        precision: torch.float16
-        enable: false           # enable or disable this model
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Baichuan2-7B-Base.jsonl
-    Qwen-7B:
-        model_id_or_path: qwen/Qwen-7B
-        revision: v1.1.8       # revision of model, default is NULL
-        precision: torch.float16
-        enable: true           # enable or disable this model      # TODO: tokenizer issue
-        template_type: default-generation
-        generation_config:
-            do_sample: true
-            max_new_tokens: 256
-            top_k: 20
-            top_p: 0.75
-            temperature: 0.3
-        output_file: registry/data/arena/answers/answer_Qwen-7B.jsonl
-
-# model of auto-reviewer
-reviews_gen:
-    enable: true
-    reviewer:
-        ref: evalscope.evaluator.reviewer.auto_reviewer:AutoReviewerGpt4
-        args:
-            model: gpt-4
-            max_tokens: 1024
-            temperature: 0
-            # pairwise comparison against baseline
-            mode: single
-            # completion parser config, default is lmsys_parser
-            fn_completion_parser: lmsys_parser
-    # target answers list to be reviewed, could be replaced by your own path: /path/to/answers.jsonl
-    target_answers: [registry/data/arena/answers/answer_chatglm3-6b.jsonl,
-                     registry/data/arena/answers/answer_Baichuan2-7B-Base.jsonl]
-    # the path to the reference answers
-    reference_file:
-    # prompt templates for auto reviewer(GPT-4)
-    prompt_file: registry/data/prompt_template/lmsys_v2.jsonl
-    # output file of auto reviewer
-    review_file: registry/data/arena/reviews/review_gpt4_single.jsonl
-    # cache file of auto reviewer
-    cache_file: registry/data/arena/reviews/review_gpt4_single.jsonl
-
-# rating results
-rating_gen:
-    enable: true
-    metrics: ['score']
-    # elo rating report file
-    report_file: registry/data/arena/reports/rating_single.csv

evalscope/registry/data/prompt_template/lmsys_v2.jsonl

@@ -1,8 +0,0 @@
-{"name": "pair-v2", "type": "pairwise", "system_prompt": "Please act as an impartial judge and evaluate the quality of the responses provided by two AI assistants to the user question displayed below. You should choose the assistant that follows the user's instructions and answers the user's question better. Your evaluation should consider factors such as the helpfulness, relevance, accuracy, depth, creativity, and level of detail of their responses. Begin your evaluation by comparing the two responses and provide a short explanation. Avoid any positional biases and ensure that the order in which the responses were presented does not influence your decision. Do not allow the length of the responses to influence your evaluation. Do not favor certain names of the assistants. Be as objective as possible. After providing your explanation, output your final verdict by strictly following this format: \"[[A]]\" if assistant A is better, \"[[B]]\" if assistant B is better, and \"[[C]]\" for a tie.", "prompt_template": "[User Question]\n{question}\n\n[The Start of Assistant A's Answer]\n{answer_a}\n[The End of Assistant A's Answer]\n\n[The Start of Assistant B's Answer]\n{answer_b}\n[The End of Assistant B's Answer]", "description": "Prompt for general questions", "category": "general", "output_format": "[[A]]"}
-{"name": "pair-v2-multi-turn", "type": "pairwise", "system_prompt": "Please act as an impartial judge and evaluate the quality of the responses provided by two AI assistants to the user questions. You should choose the assistant that follows the user's instructions and answers the user's questions better. Your evaluation should consider factors such as the helpfulness, relevance, accuracy, depth, creativity, and level of detail of their responses. You should focus on who provides a better answer to the second user question. Begin your evaluation by comparing the responses of the two assistants and provide a short explanation. Avoid any positional biases and ensure that the order in which the responses were presented does not influence your decision. Do not allow the length of the responses to influence your evaluation. Do not favor certain names of the assistants. Be as objective as possible. After providing your explanation, output your final verdict by strictly following this format: \"[[A]]\" if assistant A is better, \"[[B]]\" if assistant B is better, and \"[[C]]\" for a tie.", "prompt_template": "<|The Start of Assistant A's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant A:\n{answer_a_1}\n\n### User:\n{question_2}\n\n### Assistant A:\n{answer_a_2}\n\n<|The End of Assistant A's Conversation with User|>\n\n\n<|The Start of Assistant B's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant B:\n{answer_b_1}\n\n### User:\n{question_2}\n\n### Assistant B:\n{answer_b_2}\n\n<|The End of Assistant B's Conversation with User|>", "description": "Prompt for multi-turn general questions", "category": "general", "output_format": "[[A]]"}
-{"name": "pair-math-v1", "type": "pairwise", "system_prompt": "Please act as an impartial judge and evaluate the quality of the responses provided by two AI assistants to the user question displayed below. Your evaluation should consider correctness and helpfulness. You will be given a reference answer, assistant A's answer, and assistant B's answer. Your job is to evaluate which assistant's answer is better. Begin your evaluation by comparing both assistants' answers with the reference answer. Identify and correct any mistakes. Avoid any positional biases and ensure that the order in which the responses were presented does not influence your decision. Do not allow the length of the responses to influence your evaluation. Do not favor certain names of the assistants. Be as objective as possible. After providing your explanation, output your final verdict by strictly following this format: \"[[A]]\" if assistant A is better, \"[[B]]\" if assistant B is better, and \"[[C]]\" for a tie.", "prompt_template": "[User Question]\n{question}\n\n[The Start of Reference Answer]\n{ref_answer_1}\n[The End of Reference Answer]\n\n[The Start of Assistant A's Answer]\n{answer_a}\n[The End of Assistant A's Answer]\n\n[The Start of Assistant B's Answer]\n{answer_b}\n[The End of Assistant B's Answer]", "description": "Prompt for math questions", "category": ["math", "reasoning", "coding"], "output_format": "[[A]]"}
-{"name": "pair-math-v1-multi-turn", "type": "pairwise", "system_prompt": "Please act as an impartial judge and evaluate the quality of the responses provided by two AI assistants to the user questions. Your evaluation should consider correctness and helpfulness. You will be given reference answers, the assistant A's answers, the assistant B's answers. Your job is to determine which assistant provides correct and helpful answers to the second user question. Begin your evaluation by comparing both assistants' answers with the reference answers. Identify and correct any mistakes. Avoid any positional biases and ensure that the order in which the responses were presented does not influence your decision. Do not allow the length of the responses to influence your evaluation. Do not favor certain names of the assistants. Be as objective as possible. After providing your explanation, output your final verdict by strictly following this format: \"[[A]]\" if assistant A is better, \"[[B]]\" if assistant B is better, and \"[[C]]\" for a tie.", "prompt_template": "<|The Start of Reference Answer|>\n\n### User:\n{question_1}\n\n### Reference answer:\n{ref_answer_1}\n\n### User:\n{question_2}\n\n### Reference answer:\n{ref_answer_2}\n\n<|The End of Reference Answer|>\n\n\n<|The Start of Assistant A's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant A:\n{answer_a_1}\n\n### User:\n{question_2}\n\n### Assistant A:\n{answer_a_2}\n\n<|The End of Assistant A's Conversation with User|>\n\n\n<|The Start of Assistant B's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant B:\n{answer_b_1}\n\n### User:\n{question_2}\n\n### Assistant B:\n{answer_b_2}\n\n<|The End of Assistant B's Conversation with User|>", "description": "Prompt for multi-turn general questions", "category": ["math", "reasoning", "coding"], "output_format": "[[A]]"}
-{"name": "single-v1", "type": "single", "system_prompt": "You are a helpful assistant.", "prompt_template": "[Instruction]\nPlease act as an impartial judge and evaluate the quality of the response provided by an AI assistant to the user question displayed below. Your evaluation should consider factors such as the helpfulness, relevance, accuracy, depth, creativity, and level of detail of the response. Begin your evaluation by providing a short explanation. Be as objective as possible. After providing your explanation, you must rate the response on a scale of 1 to 10 by strictly following this format: \"[[rating]]\", for example: \"Rating: [[5]]\".\n\n[Question]\n{question}\n\n[The Start of Assistant's Answer]\n{answer}\n[The End of Assistant's Answer]", "description": "Prompt for general questions", "category": "general", "output_format": "[[rating]]"}
-{"name": "single-math-v1", "type": "single", "system_prompt": "You are a helpful assistant.", "prompt_template": "[Instruction]\nPlease act as an impartial judge and evaluate the quality of the response provided by an AI assistant to the user question displayed below. Your evaluation should consider correctness and helpfulness. You will be given a reference answer and the assistant's answer. Begin your evaluation by comparing the assistant's answer with the reference answer. Identify and correct any mistakes. Be as objective as possible. After providing your explanation, you must rate the response on a scale of 1 to 10 by strictly following this format: \"[[rating]]\", for example: \"Rating: [[5]]\".\n\n[Question]\n{question}\n\n[The Start of Reference Answer]\n{ref_answer_1}\n[The End of Reference Answer]\n\n[The Start of Assistant's Answer]\n{answer}\n[The End of Assistant's Answer]", "description": "Prompt for general questions", "category": ["math", "reasoning", "coding"], "output_format": "[[rating]]"}
-{"name": "single-v1-multi-turn", "type": "single", "system_prompt": "Please act as an impartial judge and evaluate the quality of the response provided by an AI assistant to the user question displayed below. Your evaluation should consider factors such as the helpfulness, relevance, accuracy, depth, creativity, and level of detail of the response. You evaluation should focus on the assistant's answer to the second user question. Begin your evaluation by providing a short explanation. Be as objective as possible. After providing your explanation, you must rate the response on a scale of 1 to 10 by strictly following this format: \"[[rating]]\", for example: \"Rating: [[5]]\".\n\n", "prompt_template": "<|The Start of Assistant A's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant A:\n{answer_1}\n\n### User:\n{question_2}\n\n### Assistant A:\n{answer_2}\n\n<|The End of Assistant A's Conversation with User|>", "description": "Prompt for general questions", "category": "general", "output_format": "[[rating]]"}
-{"name": "single-math-v1-multi-turn", "type": "single", "system_prompt": "Please act as an impartial judge and evaluate the quality of the response provided by an AI assistant to the user question. Your evaluation should consider correctness and helpfulness. You will be given a reference answer and the assistant's answer. You evaluation should focus on the assistant's answer to the second question. Begin your evaluation by comparing the assistant's answer with the reference answer. Identify and correct any mistakes. Be as objective as possible. After providing your explanation, you must rate the response on a scale of 1 to 10 by strictly following this format: \"[[rating]]\", for example: \"Rating: [[5]]\".\n\n", "prompt_template": "<|The Start of Reference Answer|>\n\n### User:\n{question_1}\n\n### Reference answer:\n{ref_answer_1}\n\n### User:\n{question_2}\n\n### Reference answer:\n{ref_answer_2}\n\n<|The End of Reference Answer|>\n\n\n<|The Start of Assistant A's Conversation with User|>\n\n### User:\n{question_1}\n\n### Assistant A:\n{answer_1}\n\n### User:\n{question_2}\n\n### Assistant A:\n{answer_2}\n\n<|The End of Assistant A's Conversation with User|>", "description": "Prompt for general questions", "category": ["math", "reasoning", "coding"], "output_format": "[[rating]]"}