This is a problem-set for the SherlockBench AI benchmarking system.
This problem-set is designed to test fluid intelligence. It only has three problems because
- current frontier models are extremely slow and expensive to test
- we're not testing knowledge but investigation ability so having a large problem-set seems unnecessary
Copy the source file into your checkout of the
SherlockBench-API
under src/clj/.
Then edit config.edn and add it to :problem-namespaces.
Note this problem-set contains procedurally generated problems.
To make them reproducable, make a custom problem-set and set seed. For example
"crack-lock-seeded" {:names [("irene1" "crack lock")] :seed 42}
It's just a single file.
Run the tests like this:
clj -X:test
This is Irene1v1