---
pretty_name: FTB-Core v1
configs:
- config_name: default
  data_files:
  - split: train
    path: train.jsonl.gz
  - split: validation
    path: validation.jsonl.gz
  - split: test
    path: test.jsonl.gz
  - split: test_ood
    path: test_ood.jsonl.gz
---
# FTB-Core v1

Fold2Reason project-generated spatial reasoning benchmark. The four splits contain
36,000 train, 6,000 validation, 6,000 test, and 6,000 out-of-distribution test
questions. The paper's FTB-Core evaluation uses `test` plus `test_ood`.

```python
from datasets import load_dataset
dataset = load_dataset("GENTEL-Lab/FTB-Core-v1")
```

Paper: https://huggingface.co/papers/2609.38879  
Code and evaluation protocol: https://github.com/GENTEL-lab/Fold2Reason

Source: project `dataset.zip`, SHA-256 `97ef38130fec0b7caaa323022d5b7bef620e7b26dfa42a1da4713066806dd90b`. Original compressed
JSONL files are unchanged; see `SOURCE_PROVENANCE.json` for file hashes.