New
openai/multilangbench — an evaluation suite requiring an SQLite 3.40.1 implementation
openai/multilangbench asks language models to write an SQLite 3.40.1 implementation themselves and scores the result by comparing it against real SQLite.
openai/multilangbench asks language models to write an SQLite 3.40.1 implementation themselves and scores the result by comparing it against real SQLite.
View on GitHub ↗openai/multilangbench asks language models to write an SQLite 3.40.1 implementation themselves and scores the result by comparing it against real SQLite.
No matching stories.