evalscope 0.16.2__py3-none-any.whl → 0.17.0__py3-none-any.whl

evalscope/report/utils.py

@@ -3,14 +3,45 @@ import os
 import pandas as pd
 from collections import defaultdict
 from dataclasses import asdict, dataclass, field
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Union
 
 from evalscope.metrics import macro_mean, micro_mean
-from evalscope.utils import normalize_score
-from evalscope.utils.logger import get_logger
+from evalscope.utils import get_logger
 
 logger = get_logger()
 
+ANALYSIS_PROMPT = """根据给出的json格式的模型评测结果，输出分析报告，要求如下：
+1. 报告分为 总体表现、关键指标分析、改进建议、结论 四部分
+2. 若模型有多种指标，将其分为低分、中分、高分三个部分，并列出markdown表格
+3. 只列出报告本身，不要有其他多余内容
+4. 输出报告语言为{language}
+
+```json
+{report_str}
+```
+"""
+
+
+def normalize_score(score: Union[float, dict], keep_num: int = 4) -> Union[float, dict]:
+    """
+    Normalize score.
+
+    Args:
+        score: input score, could be float or dict. e.g. 0.12345678 or {'acc': 0.12345678, 'f1': 0.12345678}
+        keep_num: number of digits to keep.
+
+    Returns:
+        Union[float, dict]: normalized score. e.g. 0.1234 or {'acc': 0.1234, 'f1': 0.1234}
+    """
+    if isinstance(score, float):
+        score = round(score, keep_num)
+    elif isinstance(score, dict):
+        score = {k: round(v, keep_num) for k, v in score.items()}
+    else:
+        logger.warning(f'Unknown score type: {type(score)}')
+
+    return score
+
 
 @dataclass
 class Subset:
@@ -74,18 +105,6 @@ class ReportKey:
     score = 'Score'
 
 
-ANALYSIS_PROMPT = """根据给出的json格式的模型评测结果，输出分析报告，要求如下：
-1. 报告分为 总体表现、关键指标分析、改进建议、结论 四部分
-2. 若模型有多种指标，将其分为低分、中分、高分三个部分，并列出markdown表格
-3. 只列出报告本身，不要有其他多余内容
-4. 输出报告语言为{language}
-
-```json
-{report_str}
-```
-"""
-
-
 @dataclass
 class Report:
     name: str = 'default_report'

evalscope/run.py

@@ -9,9 +9,9 @@ from typing import TYPE_CHECKING, List, Optional, Union
 
 from evalscope.config import TaskConfig, parse_task_config
 from evalscope.constants import DataCollection, EvalBackend
-from evalscope.utils import seed_everything
 from evalscope.utils.io_utils import OutputsStructure
 from evalscope.utils.logger import configure_logging, get_logger
+from evalscope.utils.model_utils import seed_everything
 
 if TYPE_CHECKING:
     from evalscope.models import LocalModel

evalscope/summarizer.py

@@ -7,8 +7,7 @@ from typing import List, Union
 from evalscope.config import TaskConfig, parse_task_config
 from evalscope.constants import EvalBackend
 from evalscope.report import gen_table
-from evalscope.utils import csv_to_list, get_latest_folder_path
-from evalscope.utils.io_utils import OutputsStructure, json_to_dict, yaml_to_dict
+from evalscope.utils.io_utils import OutputsStructure, csv_to_list, get_latest_folder_path, json_to_dict, yaml_to_dict
 from evalscope.utils.logger import get_logger
 
 logger = get_logger()

evalscope/utils/__init__.py

@@ -1,4 +1,65 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-from evalscope.utils.model_utils import EvalBackend
-from evalscope.utils.utils import *
+from typing import TYPE_CHECKING
+
+from .import_utils import _LazyModule
+
+if TYPE_CHECKING:
+    from .argument_utils import BaseArgument, get_supported_params, parse_int_or_float
+    from .deprecation_utils import deprecated
+    from .import_utils import get_module_path, is_module_installed
+    from .io_utils import (OutputsStructure, csv_to_jsonl, csv_to_list, dict_to_yaml, gen_hash, get_latest_folder_path,
+                           get_valid_list, json_to_dict, jsonl_to_csv, jsonl_to_list, yaml_to_dict)
+    from .logger import configure_logging, get_logger
+    from .model_utils import EvalBackend, dict_torch_dtype_to_str, fix_do_sample_warning, get_device, seed_everything
+
+else:
+    _import_structure = {
+        'argument_utils': [
+            'BaseArgument',
+            'parse_int_or_float',
+            'get_supported_params',
+        ],
+        'model_utils': [
+            'EvalBackend',
+            'get_device',
+            'seed_everything',
+            'dict_torch_dtype_to_str',
+            'fix_do_sample_warning',
+        ],
+        'import_utils': [
+            'is_module_installed',
+            'get_module_path',
+        ],
+        'io_utils': [
+            'OutputsStructure',
+            'csv_to_list',
+            'json_to_dict',
+            'yaml_to_dict',
+            'get_latest_folder_path',
+            'gen_hash',
+            'dict_to_yaml',
+            'csv_to_jsonl',
+            'jsonl_to_csv',
+            'jsonl_to_list',
+            'gen_hash',
+            'get_valid_list',
+        ],
+        'deprecation_utils': [
+            'deprecated',
+        ],
+        'logger': [
+            'get_logger',
+            'configure_logging',
+        ],
+    }
+
+    import sys
+
+    sys.modules[__name__] = _LazyModule(
+        __name__,
+        globals()['__file__'],
+        _import_structure,
+        module_spec=__spec__,
+        extra_objects={},
+    )

evalscope/utils/argument_utils.py

@@ -0,0 +1,64 @@
+import json
+from argparse import Namespace
+from inspect import signature
+
+from evalscope.utils.io_utils import json_to_dict, yaml_to_dict
+
+
+class BaseArgument:
+    """
+    BaseArgument is a base class designed to facilitate the creation and manipulation
+    of argument classes in the evalscope framework. It provides utility methods for
+    instantiating objects from various data formats and converting objects back into
+    dictionary representations.
+    """
+
+    @classmethod
+    def from_dict(cls, d: dict):
+        """Instantiate the class from a dictionary."""
+        return cls(**d)
+
+    @classmethod
+    def from_json(cls, json_file: str):
+        """Instantiate the class from a JSON file."""
+        return cls.from_dict(json_to_dict(json_file))
+
+    @classmethod
+    def from_yaml(cls, yaml_file: str):
+        """Instantiate the class from a YAML file."""
+        return cls.from_dict(yaml_to_dict(yaml_file))
+
+    @classmethod
+    def from_args(cls, args: Namespace):
+        """
+        Instantiate the class from an argparse.Namespace object.
+        Filters out None values and removes 'func' if present.
+        """
+        args_dict = {k: v for k, v in vars(args).items() if v is not None}
+
+        if 'func' in args_dict:
+            del args_dict['func']  # Note: compat CLI arguments
+
+        return cls.from_dict(args_dict)
+
+    def to_dict(self):
+        """Convert the instance to a dictionary."""
+        result = self.__dict__.copy()
+        return result
+
+    def __str__(self):
+        """Return a JSON-formatted string representation of the instance."""
+        return json.dumps(self.to_dict(), indent=4, default=str, ensure_ascii=False)
+
+
+def parse_int_or_float(num):
+    number = float(num)
+    if number.is_integer():
+        return int(number)
+    return number
+
+
+def get_supported_params(func):
+    """Get the supported parameters of a function."""
+    sig = signature(func)
+    return list(sig.parameters.keys())

evalscope/utils/import_utils.py

@@ -64,3 +64,19 @@ class _LazyModule(ModuleType):
 
     def __reduce__(self):
         return self.__class__, (self._name, self.__file__, self._import_structure)
+
+
+def is_module_installed(module_name):
+    try:
+        importlib.import_module(module_name)
+        return True
+    except ImportError:
+        return False
+
+
+def get_module_path(module_name):
+    spec = importlib.util.find_spec(module_name)
+    if spec and spec.origin:
+        return os.path.abspath(spec.origin)
+    else:
+        raise ValueError(f'Cannot find module: {module_name}')

evalscope/utils/io_utils.py

@@ -1,7 +1,9 @@
 import csv
+import hashlib
 import json
 import jsonlines as jsonl
 import os
+import re
 import yaml
 
 from evalscope.constants import DumpMode
@@ -221,7 +223,46 @@ def dict_to_json(d: dict, json_file: str):
         json.dump(d, f, indent=4, ensure_ascii=False)
 
 
-if __name__ == '__main__':
-    csv_file = 'custom_eval/text/mcq/example_val.csv'
-    jsonl_file = 'custom_eval/text/mcq/example_val.jsonl'
-    csv_to_jsonl(csv_file, jsonl_file)
+def get_latest_folder_path(work_dir):
+    from datetime import datetime
+
+    # Get all subdirectories in the work_dir
+    folders = [f for f in os.listdir(work_dir) if os.path.isdir(os.path.join(work_dir, f))]
+
+    # Get the timestamp（YYYYMMDD_HHMMSS）
+    timestamp_pattern = re.compile(r'^\d{8}_\d{6}$')
+
+    # Filter out the folders
+    timestamped_folders = [f for f in folders if timestamp_pattern.match(f)]
+
+    if not timestamped_folders:
+        print(f'>> No timestamped folders found in {work_dir}!')
+        return None
+
+    # timestamp parser
+    def parse_timestamp(folder_name):
+        return datetime.strptime(folder_name, '%Y%m%d_%H%M%S')
+
+    # Find the latest folder
+    latest_folder = max(timestamped_folders, key=parse_timestamp)
+
+    return os.path.join(work_dir, latest_folder)
+
+
+def gen_hash(name: str, bits: int = 32):
+    return hashlib.md5(name.encode(encoding='UTF-8')).hexdigest()[:bits]
+
+
+def get_valid_list(input_list, candidate_list):
+    """
+    Get the valid and invalid list from input_list based on candidate_list.
+    Args:
+        input_list: The input list.
+        candidate_list: The candidate list.
+
+    Returns:
+        valid_list: The valid list.
+        invalid_list: The invalid list.
+    """
+    return [i for i in input_list if i in candidate_list], \
+           [i for i in input_list if i not in candidate_list]

evalscope/utils/model_utils.py

@@ -1,6 +1,9 @@
+import numpy as np
 import os
+import random
+import torch
 from enum import Enum
-from typing import TYPE_CHECKING, Optional, Tuple, Union
+from typing import TYPE_CHECKING, Any, Dict, Optional, Tuple, Union
 
 if TYPE_CHECKING:
     from transformers import GenerationConfig
@@ -38,3 +41,36 @@ def get_device() -> str:
         device = 'cpu'
 
     return device
+
+
+def dict_torch_dtype_to_str(d: Dict[str, Any]) -> dict:
+    """
+        Checks whether the passed dictionary and its nested dicts have a *torch_dtype* key and if it's not None,
+        converts torch.dtype to a string of just the type. For example, `torch.float32` get converted into *"float32"*
+        string, which can then be stored in the json format.
+
+        Refer to: https://github.com/huggingface/transformers/pull/16065/files for details.
+        """
+    if d.get('torch_dtype', None) is not None and not isinstance(d['torch_dtype'], str):
+        d['torch_dtype'] = str(d['torch_dtype']).split('.')[1]
+
+    for value in d.values():
+        if isinstance(value, dict):
+            dict_torch_dtype_to_str(value)
+
+    return d
+
+
+def seed_everything(seed: int):
+    """Set all random seeds to a fixed value for reproducibility.
+
+    Args:
+        seed (int): The seed value.
+    """
+    random.seed(seed)
+    np.random.seed(seed)
+    torch.manual_seed(seed)
+    if torch.cuda.is_available():
+        torch.cuda.manual_seed_all(seed)
+        torch.backends.cudnn.deterministic = True
+        torch.backends.cudnn.benchmark = False

evalscope/version.py

@@ -1,4 +1,4 @@
 # Copyright (c) Alibaba, Inc. and its affiliates.
 
-__version__ = '0.16.2'
-__release_datetime__ = '2025-06-23 20:00:00'
+__version__ = '0.17.0'
+__release_datetime__ = '2025-07-04 17:00:00'

{evalscope-0.16.2.dist-info → evalscope-0.17.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: evalscope
-Version: 0.16.2
+Version: 0.17.0
 Summary: EvalScope: Lightweight LLMs Evaluation Framework
 Home-page: https://github.com/modelscope/evalscope
 Author: ModelScope team
@@ -17,14 +17,14 @@ Requires-Python: >=3.8
 Description-Content-Type: text/markdown
 License-File: LICENSE
 Requires-Dist: accelerate
-Requires-Dist: datasets>=3.0
+Requires-Dist: datasets==3.2.0
 Requires-Dist: immutabledict
 Requires-Dist: jieba
 Requires-Dist: jsonlines
 Requires-Dist: langdetect
 Requires-Dist: latex2sympy2-extended
 Requires-Dist: matplotlib
-Requires-Dist: modelscope[framework]
+Requires-Dist: modelscope[framework]>=1.27
 Requires-Dist: nltk>=3.9
 Requires-Dist: openai
 Requires-Dist: pandas
@@ -52,14 +52,14 @@ Requires-Dist: opencv-python; extra == "aigc"
 Requires-Dist: torchvision; extra == "aigc"
 Provides-Extra: all
 Requires-Dist: accelerate; extra == "all"
-Requires-Dist: datasets>=3.0; extra == "all"
+Requires-Dist: datasets==3.2.0; extra == "all"
 Requires-Dist: immutabledict; extra == "all"
 Requires-Dist: jieba; extra == "all"
 Requires-Dist: jsonlines; extra == "all"
 Requires-Dist: langdetect; extra == "all"
 Requires-Dist: latex2sympy2-extended; extra == "all"
 Requires-Dist: matplotlib; extra == "all"
-Requires-Dist: modelscope[framework]; extra == "all"
+Requires-Dist: modelscope[framework]>=1.27; extra == "all"
 Requires-Dist: nltk>=3.9; extra == "all"
 Requires-Dist: openai; extra == "all"
 Requires-Dist: pandas; extra == "all"
@@ -94,7 +94,7 @@ Requires-Dist: rich; extra == "all"
 Requires-Dist: sse-starlette; extra == "all"
 Requires-Dist: transformers; extra == "all"
 Requires-Dist: uvicorn; extra == "all"
-Requires-Dist: gradio==5.4.0; extra == "all"
+Requires-Dist: gradio>=5.4.0; extra == "all"
 Requires-Dist: plotly<6.0.0,>=5.23.0; extra == "all"
 Requires-Dist: diffusers; extra == "all"
 Requires-Dist: iopath; extra == "all"
@@ -102,9 +102,20 @@ Requires-Dist: omegaconf; extra == "all"
 Requires-Dist: open-clip-torch; extra == "all"
 Requires-Dist: opencv-python; extra == "all"
 Requires-Dist: torchvision; extra == "all"
+Requires-Dist: bfcl-eval; extra == "all"
+Requires-Dist: dotenv; extra == "all"
+Requires-Dist: human-eval; extra == "all"
+Requires-Dist: pytest; extra == "all"
+Requires-Dist: pytest-cov; extra == "all"
 Provides-Extra: app
-Requires-Dist: gradio==5.4.0; extra == "app"
+Requires-Dist: gradio>=5.4.0; extra == "app"
 Requires-Dist: plotly<6.0.0,>=5.23.0; extra == "app"
+Provides-Extra: dev
+Requires-Dist: bfcl-eval; extra == "dev"
+Requires-Dist: dotenv; extra == "dev"
+Requires-Dist: human-eval; extra == "dev"
+Requires-Dist: pytest; extra == "dev"
+Requires-Dist: pytest-cov; extra == "dev"
 Provides-Extra: opencompass
 Requires-Dist: ms-opencompass>=0.1.6; extra == "opencompass"
 Provides-Extra: perf
@@ -198,24 +209,33 @@ EvalScope is not merely an evaluation tool; it is a valuable ally in your model
 Below is the overall architecture diagram of EvalScope:
 
 <p align="center">
-  <img src="docs/en/_static/images/evalscope_framework.png" width="70%">
+  <img src="https://sail-moe.oss-cn-hangzhou.aliyuncs.com/yunlin/images/evalscope/doc/EvalScope%E6%9E%B6%E6%9E%84%E5%9B%BE.png" width="70%">
   <br>EvalScope Framework.
 </p>
 
 <details><summary>Framework Description</summary>
 
 The architecture includes the following modules:
-1. **Model Adapter**: The model adapter is used to convert the outputs of specific models into the format required by the framework, supporting both API call models and locally run models.
-2. **Data Adapter**: The data adapter is responsible for converting and processing input data to meet various evaluation needs and formats.
-3. **Evaluation Backend**:
-    - **Native**: EvalScope’s own **default evaluation framework**, supporting various evaluation modes, including single model evaluation, arena mode, baseline model comparison mode, etc.
-    - **OpenCompass**: Supports [OpenCompass](https://github.com/open-compass/opencompass) as the evaluation backend, providing advanced encapsulation and task simplification, allowing you to submit tasks for evaluation more easily.
-    - **VLMEvalKit**: Supports [VLMEvalKit](https://github.com/open-compass/VLMEvalKit) as the evaluation backend, enabling easy initiation of multi-modal evaluation tasks, supporting various multi-modal models and datasets.
-    - **RAGEval**: Supports RAG evaluation, supporting independent evaluation of embedding models and rerankers using [MTEB/CMTEB](https://evalscope.readthedocs.io/en/latest/user_guides/backend/rageval_backend/mteb.html), as well as end-to-end evaluation using [RAGAS](https://evalscope.readthedocs.io/en/latest/user_guides/backend/rageval_backend/ragas.html).
-    - **ThirdParty**: Other third-party evaluation tasks, such as ToolBench.
-4. **Performance Evaluator**: Model performance evaluation, responsible for measuring model inference service performance, including performance testing, stress testing, performance report generation, and visualization.
-5. **Evaluation Report**: The final generated evaluation report summarizes the model's performance, which can be used for decision-making and further model optimization.
-6. **Visualization**: Visualization results help users intuitively understand evaluation results, facilitating analysis and comparison of different model performances.
+1. Input Layer
+- **Model Sources**: API models (OpenAI API), local models (ModelScope)
+- **Datasets**: Standard evaluation benchmarks (MMLU/GSM8k, etc.), custom data (MCQ/QA)
+
+2. Core Functions
+- **Multi-backend Evaluation**
+   - Native backends: Unified evaluation for LLM/VLM/Embedding/T2I models
+   - Integrated frameworks: OpenCompass/MTEB/VLMEvalKit/RAGAS
+
+- **Performance Monitoring**
+   - Model plugins: Supports various model service APIs
+   - Data plugins: Supports multiple data formats
+   - Metric tracking: TTFT/TPOP/Stability and other metrics
+
+- **Tool Extensions**
+   - Integration: Tool-Bench/Needle-in-a-Haystack/BFCL-v3
+
+3. Output Layer
+- **Structured Reports**: Supports JSON/Tables/Logs
+- **Visualization Platforms**: Supports Gradio/Wandb/SwanLab
 
 </details>
 
@@ -230,7 +250,9 @@ Please scan the QR code below to join our community groups:
 
 ## 🎉 News
 
-- 🔥 **[2025.06.19]** Added support for the BFCL-v3 benchmark, designed to evaluate model function-calling capabilities across various scenarios. For more information, refer to the [documentation](https://evalscope.readthedocs.io/zh-cn/latest/third_party/bfcl_v3.html).
+- 🔥 **[2025.07.03]** Refactored Arena Mode: now supports custom model battles, outputs a model leaderboard, and provides battle result visualization. See [reference](https://evalscope.readthedocs.io/en/latest/user_guides/arena.html) for details.
+- 🔥 **[2025.06.28]** Optimized custom dataset evaluation: now supports evaluation without reference answers. Enhanced LLM judge usage, with built-in modes for "scoring directly without reference answers" and "checking answer consistency with reference answers". See [reference](https://evalscope.readthedocs.io/en/latest/advanced_guides/custom_dataset/llm.html#qa) for details.
+- 🔥 **[2025.06.19]** Added support for the [BFCL-v3](https://modelscope.cn/datasets/AI-ModelScope/bfcl_v3) benchmark, designed to evaluate model function-calling capabilities across various scenarios. For more information, refer to the [documentation](https://evalscope.readthedocs.io/zh-cn/latest/third_party/bfcl_v3.html).
 - 🔥 **[2025.06.02]** Added support for the Needle-in-a-Haystack test. Simply specify `needle_haystack` to conduct the test, and a corresponding heatmap will be generated in the `outputs/reports` folder, providing a visual representation of the model's performance. Refer to the [documentation](https://evalscope.readthedocs.io/en/latest/third_party/needle_haystack.html) for more details.
 - 🔥 **[2025.05.29]** Added support for two long document evaluation benchmarks: [DocMath](https://modelscope.cn/datasets/yale-nlp/DocMath-Eval/summary) and [FRAMES](https://modelscope.cn/datasets/iic/frames/summary). For usage guidelines, please refer to the [documentation](https://evalscope.readthedocs.io/en/latest/get_started/supported_dataset.html).
 - 🔥 **[2025.05.16]** Model service performance stress testing now supports setting various levels of concurrency and outputs a performance test report. [Reference example](https://evalscope.readthedocs.io/en/latest/user_guides/stress_test/quick_start.html#id3).
@@ -248,12 +270,12 @@ Please scan the QR code below to join our community groups:
 - 🔥 **[2025.03.03]** Added support for evaluating the IQ and EQ of models. Refer to [📖 Best Practices for IQ and EQ Evaluation](https://evalscope.readthedocs.io/en/latest/best_practice/iquiz.html) to find out how smart your AI is!
 - 🔥 **[2025.02.27]** Added support for evaluating the reasoning efficiency of models. Refer to [📖 Best Practices for Evaluating Thinking Efficiency](https://evalscope.readthedocs.io/en/latest/best_practice/think_eval.html). This implementation is inspired by the works [Overthinking](https://doi.org/10.48550/arXiv.2412.21187) and [Underthinking](https://doi.org/10.48550/arXiv.2501.18585).
 - 🔥 **[2025.02.25]** Added support for two model inference-related evaluation benchmarks: [MuSR](https://modelscope.cn/datasets/AI-ModelScope/MuSR) and [ProcessBench](https://www.modelscope.cn/datasets/Qwen/ProcessBench/summary). To use them, simply specify `musr` and `process_bench` respectively in the datasets parameter.
+<details><summary>More</summary>
+
 - 🔥 **[2025.02.18]** Supports the AIME25 dataset, which contains 15 questions (Grok3 scored 93 on this dataset).
 - 🔥 **[2025.02.13]** Added support for evaluating DeepSeek distilled models, including AIME24, MATH-500, and GPQA-Diamond datasets，refer to [best practice](https://evalscope.readthedocs.io/en/latest/best_practice/deepseek_r1_distill.html); Added support for specifying the `eval_batch_size` parameter to accelerate model evaluation.
 - 🔥 **[2025.01.20]** Support for visualizing evaluation results, including single model evaluation results and multi-model comparison, refer to the [📖 Visualizing Evaluation Results](https://evalscope.readthedocs.io/en/latest/get_started/visualization.html) for more details; Added [`iquiz`](https://modelscope.cn/datasets/AI-ModelScope/IQuiz/summary) evaluation example, evaluating the IQ and EQ of the model.
 - 🔥 **[2025.01.07]** Native backend: Support for model API evaluation is now available. Refer to the [📖 Model API Evaluation Guide](https://evalscope.readthedocs.io/en/latest/get_started/basic_usage.html#api) for more details. Additionally, support for the `ifeval` evaluation benchmark has been added.
-<details><summary>More</summary>
-
 - 🔥🔥 **[2024.12.31]** Support for adding benchmark evaluations, refer to the [📖 Benchmark Evaluation Addition Guide](https://evalscope.readthedocs.io/en/latest/advanced_guides/add_benchmark.html); support for custom mixed dataset evaluations, allowing for more comprehensive model evaluations with less data, refer to the [📖 Mixed Dataset Evaluation Guide](https://evalscope.readthedocs.io/en/latest/advanced_guides/collection/index.html).
 - 🔥 **[2024.12.13]** Model evaluation optimization: no need to pass the `--template-type` parameter anymore; supports starting evaluation with `evalscope eval --args`. Refer to the [📖 User Guide](https://evalscope.readthedocs.io/en/latest/get_started/basic_usage.html) for more details.
 - 🔥 **[2024.11.26]** The model inference service performance evaluator has been completely refactored: it now supports local inference service startup and Speed Benchmark; asynchronous call error handling has been optimized. For more details, refer to the [📖 User Guide](https://evalscope.readthedocs.io/en/latest/user_guides/stress_test/index.html).
@@ -572,10 +594,17 @@ Speed Benchmark Results:
 EvalScope supports custom dataset evaluation. For detailed information, please refer to the Custom Dataset Evaluation [📖User Guide](https://evalscope.readthedocs.io/en/latest/advanced_guides/custom_dataset/index.html)
 
 
-## 🏟️ Arena Mode
-The Arena mode allows multiple candidate models to be evaluated through pairwise battles, and can choose to use the AI Enhanced Auto-Reviewer (AAR) automatic evaluation process or manual evaluation to obtain the evaluation report.
+## ⚔️ Arena Mode
+
+Arena mode allows you to configure multiple candidate models and specify a baseline model. Evaluation is performed by pairwise battles between each candidate model and the baseline model, with the final output including each model's win rate and ranking. This method is suitable for comparative evaluation among multiple models, providing an intuitive reflection of each model's strengths and weaknesses. Refer to: Arena Mode [📖 User Guide](https://evalscope.readthedocs.io/en/latest/user_guides/arena.html)
 
-Refer to: Arena Mode [📖 User Guide](https://evalscope.readthedocs.io/en/latest/user_guides/arena.html)
+```text
+Model           WinRate (%)  CI (%)
+------------  -------------  ---------------
+qwen2.5-72b            69.3  (-13.3 / +12.2)
+qwen2.5-7b             50    (+0.0 / +0.0)
+qwen2.5-0.5b            4.7  (-2.5 / +4.4)
+```
 
 ## 👷‍♂️ Contribution
 
@@ -601,7 +630,7 @@ EvalScope, as the official evaluation tool of [ModelScope](https://modelscope.cn
 - [ ] Distributed evaluating
 - [x] Multi-modal evaluation
 - [ ] Benchmarks
-  - [ ] GAIA
+  - [x] BFCL-v3
   - [x] GPQA
   - [x] MBPP