Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark
Paper • 2501.05444 • Published • 3
Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.
EMMA Stone is a reduced version of the EMMA (Enhanced MultiModal reAsoning) benchmark with 8 samples per subject category, designed for quick testing and development.
This dataset contains:
from datasets import load_dataset
# Load specific subject
chemistry_data = load_dataset("winvswon78/emma_stone", "Chemistry", split="test")
math_data = load_dataset("winvswon78/emma_stone", "Math", split="test")
coding_data = load_dataset("winvswon78/emma_stone", "Coding", split="test")
physics_data = load_dataset("winvswon78/emma_stone", "Physics", split="test")
# Load all subjects combined
all_data = load_dataset("winvswon78/emma_stone", "All", split="test")
# Verify the dataset
print(f"Chemistry samples: {len(chemistry_data)}")
print(f"Math samples: {len(math_data)}")
print(f"Coding samples: {len(coding_data)}")
print(f"Physics samples: {len(physics_data)}")
print(f"All samples: {len(all_data)}")
print(f"Subject distribution in All: {all_data['subject']}")
If you encounter issues with the config names, you can also load the data directly:
from datasets import Dataset
import pandas as pd
# Load specific subject directly
chemistry_df = pd.read_parquet("hf://datasets/winvswon78/emma_stone/Chemistry/test-00000-of-00001.parquet")
chemistry_dataset = Dataset.from_pandas(chemistry_df)
# Load all subjects
all_df = pd.read_parquet("hf://datasets/winvswon78/emma_stone/All/test-00000-of-00001.parquet")
all_dataset = Dataset.from_pandas(all_df)
This is a sampled version of the original EMMA benchmark targeting organic multimodal reasoning across mathematics, physics, chemistry, and coding. EMMA tasks demand advanced cross-modal reasoning that cannot be solved by thinking separately in each modality.
The dataset is provided in jsonl format and contains the following attributes:
{
"pid": [string] Problem ID, e.g., “math_1”,
"question": [string] The question text,
"options": [list] Choice options for multiple-choice problems. For free-form problems, this could be a 'none' value,
"answer": [string] The correct answer for the problem,
"image_1": [image] ,
"image_2": [image] ,
"image_3": [image] ,
"image_4": [image] ,
"image_5": [image] ,
"solution": [string] The detailed thinking steps required to solve the problem,
"subject": [string] The subject of data, e.g., “Math”, “Physics”...,
"task": [string] The task of the problem, e.g., “Code Choose Vis”,
"category": [string] The category of the problem, e.g., “2D Transformation”,
"source": [string] The original source dataset of the data, e.g., “math-vista”. For handmade data, this could be “Newly annotated” ,
"type": [string] Types of questions, e.g., “Multiple Choice”, “Open-ended”,
"context": [string] Background knowledge required for the question. For problems without context, this could be a 'none' value,
}
@misc{hao2025mllmsreasonmultimodalityemma,
title={Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark},
author={Yunzhuo Hao and Jiawei Gu and Huichen Will Wang and Linjie Li and Zhengyuan Yang and Lijuan Wang and Yu Cheng},
year={2025},
eprint={2501.05444},
archivePrefix={arXiv},
primaryClass={cs.CV},
url={https://arxiv.org/abs/2501.05444},
}