Post-Launch Dataset Cleanup: Deprecating Benchmark Datasets in the Spirit of Openness

Key Info

Nearly two weeks after launch, the team is improving usability of its released datasets while deprecating all datasets used for internal or public evaluation.

Highlights

• Datasets released alongside the launch are being made easier to work with
• Despite the openness push, the author remains anti-public benchmarks
• All evaluation datasets, internal or public, are being deprecated