mlsanityData to 7 Oct 2026

Does the Proof Prove It That Way? Faithful Formalization of Elements Proofs

arXiv 2608.15432 ↗First posted 15 Aug 2026cs.AIcited by 1

Not on the map. Why?

Abstract (arXiv)

In formal verification, both the autoformalization of statements and automated proof search have been studied extensively. While automated proof search can produce a formal proof that compiles, the generated proof does not necessarily reflect how the natural-language argument arrives at its conclusion--a property we refer to as faithfulness. With faithfully formalized proofs, one can check the reasoning behind a human- or AI-written argument, and assist mathematicians in formalizing their proof sketches. However, it is particularly challenging due to misalignment of formal proof tactics and natural language reasoning. In this work, we rigorously describe a set of five necessary conditions a faithful formal proof must satisfy, and introduce Pistis, an agentic, oracle-guided proof search that produces formal Lean proofs that satisfy them. At its core is a novel faithfulness-preserving divide-and-conquer search, which we name OrderDecompose, that tracks citation dependencies and blocks unfaithful shortcuts, paired with a refutation search, that surfaces gaps and errors in the natural language proof source. OrderDecompose completes proofs that baselines cannot close even within a 12-hour budget, and its artifacts compile over 33× as fast as prior work's. We apply Pistis on the first three books of Euclid's Elements, producing high-quality artifacts containing faithful formal proofs. Under a blinded human study and an LLM-as-a-judge protocol on rigorous rubrics, Pistis-generated proofs are favored over prior works--2.89× and 5.2× as often by human reviewers and the LLM judge, respectively. It further uncovers gaps in Euclid's proofs and their translation, and can accept or refute natural language proofs written by humans or AI, demonstrating that faithful formalization is useful as a proof-checking tool.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Builds on 13 of its 48 references that are in the map. First posted on arXiv 15 Aug 2026.

Builds on

13 of 48 references.

References · 13

  1. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Promptingcited by 1,081
  2. 2024OpenHands: An Open Platform for AI Software Developers as Generalist Agentscited by 956
  3. 2019Neural Probabilistic Logic Programming in DeepProbLogcited by 389
  4. 2021MiniF2F: a cross-system benchmark for formal Olympiad-level mathematicscited by 380
  5. 2023ProofNet: Autoformalizing and Formally Proving Undergraduate-Level Mathematicscited by 180
  6. 2019The Lean mathematical librarycited by 127
  7. 2023Scallop: A Language for Neurosymbolic Programmingcited by 63
  8. 2024A Survey on Deep Learning for Theorem Provingcited by 55
  9. 2015A formal proof of the Kepler conjecturecited by 42
  10. 2024FormalAlign: Automated Alignment Evaluation for Autoformalizationcited by 19
  11. 2025ProofFlow: A Dependency Graph Approach to Faithful Proof Autoformalizationcited by 18
  12. 2026Goedel-Code-Prover: Hierarchical Proof Search for Open State-of-the-Art Code Verificationcited by 10
1 more
  1. 2026Theory-Scale Auto-Formalization of Logics for Computer Sciencecited by 3

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

Similar, not cited

Close in wording, no citation either way

  1. 2026Pseudo-Formalization for Automatic Proof Verification
  2. 2025Translating Informal Proofs into Formal Proofs Using a Chain of States
  3. 2026Automated Conjecture Resolution with Formal Verification
  4. 2025ProofBridge: Auto-Formalization of Natural Language Proofs in Lean via Joint Embeddings
  5. 2024Rango: Adaptive Retrieval-Augmented Proving for Automated Software Verification
  6. 2025DRIFT: Decompose, Retrieve, Illustrate, then Formalize Theorems
  7. 2026Do LLMs Game Formalization? Evaluating Faithfulness in Logical Reasoning
  8. 2023LeanDojo: Theorem Proving with Retrieval-Augmented Language Models