giovannidemuri/mopsa-set-eval-data-debug

MopsaSetEval Debug

CruxEval-derived benchmark where concrete arg* inputs are rewritten into Mopsa abstract inputs.

For each example we store:

• upper_bound: a sound Mopsa over-approximation
• lower_bound: a dynamic under-approximation built from generated concrete test cases

A prediction is considered correct when it stays inside the upper_bound and contains the lower_bound.

The build summary is stored in mopsa_set_eval_summary.json.