Back to datasets
Dataset assetOpen Source CommunityMedical Visual Question AnsweringImage to Text
BUAADreamer/llava-med-zh-instruct-60k
This Chinese dataset is translated from llava‑med, built using the Qwen1.5‑14B‑Chat model, and contains 60 k medical visual instruction data points. Features include a messages‑and‑images structure: messages consist of role and content fields; images are sequences. The dataset provides a training split of 56,649 samples (size: 6.66 GB) with a download size of 6.57 GB. Task categories are visual question answering and image‑to‑text. The language is Chinese, tags involve medical and biology, and the scale lies between 10 K and 100 K.
Source
hugging_face
Created
Nov 28, 2025
Updated
May 21, 2024
Signals
187 views
Availability
Linked source ready
Overview
Dataset description and usage context
Dataset Overview
Basic Information
- License: Apache‑2.0
- Language: Chinese
- Tags: Medical, Biology, llama‑factory
- Size Category: 10K < size < 100K
Dataset Content
- Features:
- messages:
- role: string
- content: string
- images: image sequences
- messages:
Dataset Split
- Training set:
- Sample count: 56,649
- Data size: 6,664,412,158.42 bytes
- Download size: 6,567,484,534 bytes
Task Types
- Visual Question Answering
- Image‑to‑Text
Configuration
- Default configuration:
- Data files:
- Split: train
- Path: data/train-*
- Data files:
Need downstream help?
Pair the dataset with AI analysis and content workflows.
Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.