IBM/AI-Descartes — reverse-engineered prompt

I want a runnable version of this AI Descartes project that makes the research easy to try.

Please set it up so it can take the included datasets and background theory, run the reasoning part together with the symbolic regression part, and recover simple scientific laws from the data. Wire up the built in examples so I can reproduce the Kepler law, Einstein time dilation, and Langmuir adsorption demos, and make it easy to switch to one of the other provided problems too.

Make the whole thing simple to use, with one clear starting command, sensible defaults, and short instructions that explain what files go in, what results come out, and where outputs are saved. If there are missing steps or old setup details, look up the paper and any current docs online and fill in the gaps.

The goal is that someone can clone the repo, run it without a lot of manual setup, and quickly see the system rediscover equations from the bundled data.