Coverage for src / local_deep_research / benchmarks / datasets.py: 0%

2 statements  

« prev     ^ index     » next       coverage.py v7.12.0, created at 2026-01-11 00:51 +0000

1""" 

2Dataset handling for benchmarks. 

3 

4This is a legacy module that provides backwards compatibility with the old 

5dataset handling functions. New code should use the classes in the datasets 

6package directly. 

7 

8Notes on BrowseComp dataset: 

9- BrowseComp data is encrypted with the canary string used as the decryption key 

10- The decrypt() function handles decrypting both problems and answers 

11- For some examples where standard decryption doesn't work, we use additional methods: 

12 1. Try using various parts of the canary string as the key 

13 2. Try using hardcoded keys that are known to work 

14 3. Use a manual mapping for specific encrypted strings that have been verified 

15""" 

16 

17# Re-export the get_available_datasets function 

18# Re-export the default dataset URLs 

19from .datasets import DEFAULT_DATASET_URLS, get_available_datasets, load_dataset 

20 

21# Re-export the load_dataset function 

22__all__ = ["DEFAULT_DATASET_URLS", "get_available_datasets", "load_dataset"]