Benchmarks aim to measure the performance of an agent for a specific problem.
They are pre-set experiments, with a pre-defined list environments to run on, and pre-set evaluation metrics.
What changes, and also the input to a benchmark, is the Agent.
Benchmarks aim to measure the performance of an agent for a specific problem.
They are pre-set experiments, with a pre-defined list environments to run on, and pre-set evaluation metrics.
What changes, and also the input to a benchmark, is the
Agent.