terminalbench_native
diapason.evals.datasets.terminalbench_native
¶
TerminalBench Native dataset — loads from the terminal-bench pip package (v2 API).
Agentic benchmark using the native terminal-bench SDK for task loading and test-based evaluation.
TerminalBenchNativeDataset
¶
Bases: DatasetProvider
TerminalBench using the native terminal-bench pip package (v2 API).