---
pretty_name: FoldBench334 processed snapshot
configs:
- config_name: default
  data_files:
  - split: test
    path: targets_bb4q10.jsonl
---
# FoldBench334 processed snapshot

The exact 334-protein BB4Q10 target records used for Fold2Reason structural
readout. `target_manifest.json` describes the evaluated target cohort. This is
a processed snapshot for the paper, not a replacement for the official FoldBench.

```python
from datasets import load_dataset
targets = load_dataset("GENTEL-Lab/FoldBench334", split="test")
```

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

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