Sign in to view source links and access this dataset
Description
ETCHR GRPO-10K is a dataset of 10,000 multimodal samples created by internlm for enhancing model editing capabilities. It contains five specific tasks: Fine-grained Perception, Chart Understanding, Maze Solving, Jigsaw Puzzle, and Spatial Understanding. Each sample includes an image to be edited and an editing instruction.
Use Cases
Training instruction-following models for fine-grained image editing based on the described image-editing pairs.
Benchmarking model performance on spatial reasoning tasks like maze solving and jigsaw puzzles.
Enhancing chart understanding capabilities in multimodal AI systems.
Developing models with improved fine-grained visual perception.
Strengths
Contains 10,000 samples, providing a substantial training corpus.
Focuses on five distinct, challenging tasks for multimodal AI.
Created by internlm, a known organization in the AI research space.
Limitations
Column-level documentation is absent; field semantics must be inferred after download.
Row count is known, but other details like file formats and sample data are unavailable.
Last updated 2026-05-22 13:08:27; freshness should be verified.
Provenance
Source
internlm via Hugging Face.
Freshness
Last updated 2026-05-22 13:08:27.
License is unknown; users should verify terms before use.