A recent study shows that PPAD, the class of problems tied to finding Nash equilibria, is hard on average in the random oracle model. The analysis uses a setting called Romantica, where all participants can query a random function. In this setting, PPAD—and the intersection of PPAD and PLS, called EoPL—is average‑case hard.

In contrast, the subclass UEoPL can be solved by an unbounded algorithm that makes only polynomially many oracle queries. The method works by following successor pointers in a directed acyclic graph: starting from node (1,1), the algorithm traverses successors row by row until it reaches a sink or a terminal row. This simple traversal guarantees a solution with only polynomial effort.

The results illustrate a sharp difference between EoPL and UEoPL in Romantica: while EoPL remains hard, UEoPL is efficiently solvable. This sharp contrast resolves the black‑box hardness landscape for major TFNP classes within the random oracle model.

However, UEoPL’s tractability does not rely on cryptographic hardness assumptions; it relies solely on the random oracle setting. The study notes that stronger assumptions, such as average‑case hardness of #SAT, are required to prove hardness for other subclasses.

These findings map the boundaries between average‑case hardness and solvability in the random oracle framework, offering new insights into the structure of total NP search problems.