PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
March 17, 20260 citationsOpen Access

The Propositional Proof System of APX₁: Formalizing Razborov-Smolensky Below APC₁

View Full Paper
KMkiye mateus

Key Points

  • This research aims to clarify the proof-theoretic strength of APX1 in establishing lower bounds for certain circuit classes.
  • Examined proof complexity related to constant-depth circuits with modular counting gates.
  • Proved that APX1 is sufficient to show certain circuit lower bounds.
  • Provided a decomposition of the proof utilizing algebraic techniques and probabilistic arguments.
  • Confirmed that APX1 can prove MODq is not in AC0[p]d.
  • Showed that algebraic dimension-counting can be formalized in PV1.
  • Introduced a new system, EF(P), for propositional translation of APX1.

Abstract

We study the proof complexity of constant-depth circuit lower bounds with modularcounting gates in bounded arithmetic. While it is known that the theory PV1 proves AC0 lowerbounds and APC1 proves lower bounds for AC0p circuits, the exact proof-theoretic strengthrequired for the Razborov–Smolensky lower bound has remained an open question. We showthat APX1, a theory introduced by Chen, Li, Oliveira, and Williams CLOW26 capturingapproximate counting, is sufficient. Specifically, we prove that APX1 ⊢ MODq /∈ AC0pd byproviding a sharp decomposition of the proof: the algebraic dimension-counting argumentformalizes in PV1, while the probabilistic approximation step requires only the first-momentmethod available in APX1, avoiding full Chernoff bounds. Furthermore, we introduceEF(P)—Extended Frege augmented with approximate counting oracle axioms—as the naturalpropositional translation of APX1, a system which, to the best of our knowledge, is absentfrom existing proof complexity literature. We conclude with motivated conjectures regardingthe Additivity Gap, the Proof Analysis Problem for EF(P), and model-theoretic obstructionsto propositional reflection.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

kiye mateus (2026) studied this question.

synapsesocial.com/papers/69b8f162deb47d591b8c656chttps://doi.org/10.5281/zenodo.19038711
Ask AI
Helpful
Bookmark
Share
View Full Paper