CresOWLve: Benchmarking Creative Problem-Solving Over Real-World Knowledge

Develop a benchmark runner for 'Creative Problem-Solving' that goes beyond simple brainteasers. This provides a standardized way to test if agents can actually synthesize new ideas.

Suggested repo: CresOWLve-Bench

"Test if your AI can actually solve new problems, or if it's just memorizing benchmarks."

Estimated effort: 40h