ELEMENTARY_MATH / README.md
ShinDJ's picture
Update README.md
6fe1837 verified
metadata
dataset_info:
  features:
    - name: index
      dtype: int64
    - name: image
      dtype: image
    - name: image_path
      dtype: string
    - name: question
      dtype: string
    - name: answer
      dtype: string
    - name: split
      dtype: string
  splits:
    - name: test
      num_bytes: 42797851
      num_examples: 448
  download_size: 37624653
  dataset_size: 42797851
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*

초등 수학 멀티모달 데이터

Evaluation toolkit
GitHub

  • 구성: 공간 지각, 패턴 인식, 논리적 사고 등 초등학생 수준의 수학적 사고력을 요구하는 이미지와 한국어 질의문 및 답변으로 구성된 벤치마크 데이터셋
  • 데이터 수량: 질의/답변 데이터 448개
  • 특징: 단순 계산이 아닌 시각적 추론이 필요한 문제(ex. 여러 방향에서 본 블록 개수 추론, 도형 회전 및 변환 등)로, 이미지 기반 공간 인지 및 논리적 사고 능력 평가
  • 목적: 시각-언어 모델의 초등 수준 수학적 사고력, 시각적 추론 능력, 공간 지각 능력을 종합적으로 평가하기 위한 벤치마크 데이터셋

Evaluation Results

Scale Model Elementary school
2B ~ 3B Qwen2-VL-2B 6.25
Qwen2.5-VL-3B 9.15
InternVL3-2B 7.81
InternVL3.5-2B 7.14
7B ~ 8B Qwen2-VL-7B 10.49
Qwen2.5-VL-7B 17.85
InternVL3-8B 13.39
InternVL3.5-8B 16.29
InternVL3.5-8B (Fast Thinking) 13.39
InternVL3.5-8B (Long Thinking) 18.97
11B ~ 14B Llama-Vision-11B 2
Gemma3-12B 13.16
InternVL3-14B 22.99
InternVL3.5-14B 19.41
InternVL3.5-14B (Fast thinking) 24.77
InternVL3.5-14B (Long thinking) 25.89
Commercial Gemini2.5-Flash 38.39
GPT-4.1 27

Acknowledgements

본 데이터셋은 2023년도 정부(과학기술정보통신부)의 재원으로 정보통신기획평가원의 지원을 받아 수행된 연구임 (과제번호: RS-2023-00216011, 사람처럼 개념적으로 이해/추론이 가능한 복합인공지능 원천기술 연구)

This work was supported by Institute of Information & Communications Technology Planning & Evaluation (IITP) grant funded by the Korea government (MSIT) (RS-2023-00216011, Development of artificial complex intelligence for conceptually understanding and inferring like human)