Skip to content

Dataset · lmms-lab-encoder

Benchmark for referring expression comprehension in images

lmms-lab-encoder/RefCOCO

Formatted version of RefCOCO for evaluating multimodal models in the lmms-eval pipeline.

What's inside

Contains images, text, and timeseries data drawn from the classic RefCOCO dataset, totaling 10K to 100K examples. It is formatted specifically for quick, one-click model evaluation via the lmms-eval framework.

What you can do with it

  • Evaluating large multimodal models (LMMs) on grounding text descriptions to image objects
  • Comparing models under a unified standard within lmms-eval
Size
10K–100K rows
Downloads this month
7.3K
Likes
40

How to load it

from datasets import load_dataset

ds = load_dataset("lmms-lab-encoder/RefCOCO")