chimera bench

Run the continuous-evolution benchmark on a demo task set. Requires a key.

Options

--limitintdefault: 0
Limit number of demo tasks (0 = all).
--model, -mstr
Override the model slug.
--fuseboolean
Use the fusion engine as the solver.
--chainboolean
Run the stateful chained benchmark (error propagation).
--hardboolean
Use the hard suite (traps / propagating chain).
--roundsintdefault: 1
Re-run the suite N times; report stagnation + cost trend across rounds.