Explicit Proofs and The Flip

This article describes a formal strategy of geometric complexity theory (GCT) to resolve the {\em self referential paradox} in the $P$ vs. $NP$ and related problems. The strategy, called the {\em flip}, is to go for {\em explicit proofs} of these problems. By an explicit proof we mean a proof that constructs proof certificates of hardness that are easy to verify, construct and decode. The main result in this paper says that (1) any proof of the arithmetic implication of the $P$ vs. $NP$ conjecture is close to an explicit proof in the sense that it can be transformed into an explicit proof by proving in addition that arithmetic circuit identity testing can be derandomized in a blackbox fashion, and (2) stronger forms of these arithmetic hardness and derandomization conjectures together imply a polynomial time algorithm for a formidable explicit construction problem in algebraic geometry. This may explain why these conjectures, which look so elementary at the surface, have turned out to be so hard.

💡 Research Summary

The paper presents a formal strategy, called the “flip,” within the framework of Geometric Complexity Theory (GCT) to address the self‑referential paradox inherent in the P versus NP problem and its related conjectures. The paradox concerns the intuition that if discovering solutions is hard, then proving hardness should also be hard, raising the possibility that a proof of P ≠ NP could be independent of set‑theoretic axioms. GCT proposes to resolve this by converting the usual hardness statements into “explicit proofs” that provide short, efficiently verifiable, and constructible certificates of hardness—called obstructions.

The authors first define a trivial obstruction: an exponentially large table that lists, for every small circuit, a counterexample input on which the circuit fails to compute the target NP‑complete function (e.g., SAT). This table is both exponentially large to store and exponentially hard to verify, mirroring the original non‑uniform P vs NP formulation. The flip strategy replaces such trivial obstructions with a family of short strings (obstructions) satisfying five properties (F0–F4): they are of polynomial length, easy to decode into a small set of inputs, easy to verify, easy to construct, and guaranteed to exist for all relevant parameters.

Two central technical assumptions enable the flip. First, arithmetic circuit identity testing (PIT) can be derandomized in a black‑box fashion, i.e., BPP ⊆ P for PIT. This is a standard hardness‑vs‑randomness assumption. Second, a specific arithmetic function E(X), defined as a sum of determinants over a collection of matrices, is conjectured to be hard for small arithmetic circuits over fields of characteristic zero (or large finite fields). E(X) encodes an NP‑complete problem (testing whether E(X)=0) and serves as the arithmetic analogue of the Boolean NP‑complete function.

Under these assumptions, the authors prove the “Flip Theorem” (Theorems 4.2 and 4.3): any proof of the arithmetic version of P vs NP can be transformed into an explicit proof. The transformation proceeds by using the deterministic PIT algorithm to verify that a candidate circuit does not compute E(X), then extracting from this verification a short obstruction string. Decoding the string yields a polynomial‑size set Sₙ,ₘ of inputs such that every circuit of size ≤ m fails on at least one input in Sₙ,ₘ. Thus the existence of a hard‑function proof is turned into an algorithm that efficiently produces explicit hardness certificates.

A stronger result, “Flip Theorem 9.2,” shows that if stronger arithmetic hardness and derandomization conjectures hold, then a formidable explicit construction problem in algebraic geometry—namely, constructing explicit defining equations for certain high‑dimensional projective varieties—can be solved in polynomial time. This bridges the gap between lower‑bound arguments (hardness) and upper‑bound constructions (explicit geometric objects), confirming the central GCT philosophy that symmetry‑based characterizations of hard functions can be leveraged to produce concrete algebraic objects.

The paper also extends the flip to the Boolean setting (Theorem 10.5). Assuming NP ⊈ P/poly and a derandomized Boolean circuit identity test, one can similarly obtain short Boolean obstructions, demonstrating that the flip is a universal meta‑strategy applicable across computational models.

Overall, the work argues that the flip is not a mere reformulation of the original problem but a genuine reduction from an exponential‑time verification task to a polynomial‑time one. It highlights the interplay between hardness‑vs‑randomness, symmetry‑based function characterizations, and explicit algebraic geometry, suggesting new avenues for research such as weakening the PIT derandomization assumption, exploring alternative hard functions, and applying the flip to other complexity separations (e.g., #P vs. FP).