Paper Detail

Simplifying Safety Proofs with Forward-Backward Reasoning and Prophecy

Eden Frenkel, Kenneth L. McMillan, Oded Padon, Sharon Shoham

arxiv Score 4.3

Published 2026-04-16 · First seen 2026-04-17

General AI

Abstract

We propose an incremental approach for safety proofs that decomposes a proof with a complex inductive invariant into a sequence of simpler proof steps. Our proof system combines rules for (i) forward reasoning using inductive invariants, (ii) backward reasoning using inductive invariants of a time-reversed system, and (iii) prophecy steps that add witnesses for existentially quantified properties. We prove each rule sound and give a construction that recovers a single safe inductive invariant from an incremental proof. The construction of the invariant demonstrates the increased complexity of a single inductive invariant compared to the invariant formulas used in an incremental proof, which may have simpler Boolean structures and fewer quantifiers and quantifier alternations. Under natural restrictions on the available invariant formulas, each proof rule strictly increases proof power. That is, each rule allows to prove more safety problems with the same set of formulas. Thus, the incremental approach is able to reduce the search space of invariant formulas needed to prove safety of a given system. A case study on Paxos, several of its variants, and Raft demonstrates that forward-backward steps can remove complex Boolean structure while prophecy eliminates quantifiers and quantifier alternations.

Workflow Status

Review status
pending
Role
unreviewed
Read priority
later
Vote
Not set.
Saved
no
Collections
Not filed yet.
Next action
Not filled yet.

Reading Brief

No structured notes yet. Add `summary_sections`, `why_relevant`, `claim_impact`, or `next_action` in `papers.jsonl` to enrich this view.

Why It Surfaced

No ranking explanation is available yet.

Tags

No tags.

BibTeX

@article{frenkel2026simplifying,
  title = {Simplifying Safety Proofs with Forward-Backward Reasoning and Prophecy},
  author = {Eden Frenkel and Kenneth L. McMillan and Oded Padon and Sharon Shoham},
  year = {2026},
  abstract = {We propose an incremental approach for safety proofs that decomposes a proof with a complex inductive invariant into a sequence of simpler proof steps. Our proof system combines rules for (i) forward reasoning using inductive invariants, (ii) backward reasoning using inductive invariants of a time-reversed system, and (iii) prophecy steps that add witnesses for existentially quantified properties. We prove each rule sound and give a construction that recovers a single safe inductive invariant fr},
  url = {https://arxiv.org/abs/2604.15266},
  keywords = {cs.LO, cs.PL},
  eprint = {2604.15266},
  archiveprefix = {arXiv},
}

Metadata

{}