multabench/core-img-cls-petfinder

MUL_IMAGE_PETFINDER

Part of MulTaBench, a benchmark for multimodal tabular learning.

Paper: MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Image · Leaderboard: multabench/leaderboard

• Target: AGE_BIN
• Rows: 14652
• Original source: https://www.kaggle.com/c/petfinder-adoption-prediction

Load the benchmark splits with multabench.load_split("MUL_IMAGE_PETFINDER", fold=0).

Column	Type
Type	int
Name	string
Breed1	string
Breed2	string
Gender	int
Color1	categorical
Color2	categorical
Color3	categorical
MaturitySize	int
FurLength	int
Vaccinated	int
Dewormed	int
Sterilized	int
Health	int
Quantity	int
Fee	int
State	categorical
VideoAmt	int
Description	string
PhotoAmt	float
AdoptionSpeed	categorical
Pet Image	image path (inside images-*.zip)
AGE_BIN	string

Citation

@misc{arazi2026multabenchbenchmarkingmultimodaltabular,
  title={MulTaBench: Benchmarking Multimodal Tabular Learning with Text and Image},
  author={Alan Arazi and Eilam Shapira and Shoham Grunblat and Mor Ventura and Elad Hoffer and Gioia Blayer and David Holzmüller and Lennart Purucker and Gaël Varoquaux and Frank Hutter and Roi Reichart},
  year={2026},
  eprint={2605.10616},
  archivePrefix={arXiv},
  primaryClass={cs.LG},
  url={https://arxiv.org/abs/2605.10616},
}