Skip to content

Dataset · Rapidata

LLM 3D code generation benchmark with human ratings

Rapidata/text-to-3d-with-llm-code

44,482 examples and over 780,000 human ratings to test LLM capability in creating 3D objects via code.

What's inside

The benchmark from Rapidata contains 44,482 rows with images and English prompts in parquet format, along with 780,424 individual real-human votes. The data was collected using the Rapidata Python SDK for pairwise model performance comparison and leaderboard construction.

What you can do with it

  • Evaluating and comparing LLMs in the task of generating 3D scenes via programming code
  • Training reward models (RLHF) based on human preferences
  • Calibrating automated visual quality evaluation systems for 3D generations
Tasks
Text → video
Languages
English
Size
10K–100K rows
Downloads this month
7.7K
Likes
51

How to load it

from datasets import load_dataset

ds = load_dataset("Rapidata/text-to-3d-with-llm-code")