FIPO-Prover: Formalization-Oriented Informal Proof Optimization for Efficient Formal Theorem Proving

1NLP2CT Lab, University of Macau, 2School of Data Science and Information Engineering, Guizhou Minzu University
*Equal contribution. Correspondence to Derek F. Wong <derekfw@um.edu.mo>
FIPO-Prover Pipeline

FIPO-Prover closes the informal-to-formal gap by searching and optimizing Formalization-oriented Informal Proof (FIP) states under Isabelle verifier feedback.

Conceptual Comparison

FIPO-Prover searches over proof directions in an evolving FIP graph and optimizes intermediate FIP states before projecting to formal Isabelle proofs. This is more efficient than exhaustively searching the full formal-proof space or the full informal-proof space.

Conceptual comparison of proof-search paradigms

Conceptual comparison of formal-based search, informal-based search, and FIPO-Prover's formalization-oriented informal-proof search.

Key Results at a Glance

Abstract

Informal proofs provide a natural interface for neural theorem proving, allowing large language models to reason in natural language before producing formal proof scripts. However, a persistent informal-to-formal gap remains: mathematically plausible reasoning may contain steps, dependencies, or intermediate claims that do not align with what a formalizer such as Isabelle can verify.

We introduce FIPO-Prover, a framework for Formalization-Oriented Informal Proof Optimization that represents informal proofs as structured Formalization-oriented Informal Proof (FIP) states organized by step-level causal pairs. FIPO-Prover performs verification-guided graph search over FIP states, combining local revision, causal-pair expansion, branch pruning, upstream causal backtracking, and memory-augmented search to make informal reasoning more projection-friendly.

On MiniF2F-Test, FIPO-Prover achieves 73.77% pass@1 and 80.33% pass@3, outperforming direct LLM formalization baselines and prior search-based frameworks. Our analysis shows that these gains arise not from additional sampling alone, but from optimizing intermediate proof states so that model-generated reasoning becomes more compatible with formal verification.

Methodology

FIPO-Prover formulates informal-to-formal theorem proving as verification-guided search over structured informal proof states. Given a theorem statement, the system constructs a Formalization-oriented Informal Proof (FIP) state and iteratively optimizes an evolving FIP search graph under Isabelle feedback.

FIP State Representation

A FIP-state is a structured informal proof designed for reliable projection, diagnosis, repair, and optimization. Each step is represented as a causal pair (cause, effect), and dependencies form a directed graph. This makes the proof locally diagnosable: when projection to Isabelle fails, verifier feedback can be localized and traced upstream.

Recovery Transitions

FIPO-Prover implements a set of verifier-feedback-driven recovery transitions:

  • Root FIP-state Initialization: rewrites raw informal proofs into explicit, formalization-oriented FIP states.
  • Local FIP-state Revision: updates causal pairs and dependencies inside the current proof state.
  • Projected-Candidate Repair: fixes local artifacts in the projected Isabelle candidate.
  • Causal-Pair Expansion: introduces missing sub-lemmas or intermediate claims.
  • Upstream Causal Backtracking: propagates blame to upstream causal pairs and regenerates the branch.
  • Branch Pruning & Memory-Augmented Search: avoid redundant exploration and repeated errors.
FIP Graph Search

FIP graph search with error routing, blame-based backtracking, and causal-pair expansion.

Main Results

We evaluate FIPO-Prover on the Isabelle split of MiniF2F (488 high-school competition-level problems, 244 test / 244 valid). The system is verified step-by-step through PISA/Isabelle2022; a theorem counts as solved only when PISA reports a finished proof state.

MiniF2F Performance

Category Method Metric MiniF2F-Test MiniF2F-Valid
ClassicalSledgehammer + heuristicspass@120.9%18.0%
ClassicalThorpass@129.9%28.3%
ClassicalThor + expert iterationpass@135.2%37.3%
ClassicalThor + Magnushammerpass@137.3%36.9%
Direct LLMGemini 3.1 Propass@154.51%
Direct LLMGemini 3 Flashpass@143.85%
Direct LLMGPT-5-Codexpass@132.79%
Direct LLMDeepSeek-V3.2-Exp-Reasonerpass@138.52%
FrameworkDSPpass@10039.3%42.6%
FrameworkPOETRYpass@142.2%42.2%
FrameworkLEGO-Proverpass@10050.0%57.0%
FrameworkSubgoalXLpass@1638456.1%61.9%
FrameworkHybridProverpass@12859.4%
FrameworkFIPO-Prover (Ours)pass@173.77%81.55%
FrameworkFIPO-Prover (Ours)pass@380.33%

Results are reported on MiniF2F Isabelle test and validation splits, with each baseline kept under its originally reported sampling budget. FIPO-Prover sets a new state of the art on the Isabelle split under strict pass@1 verification.

Ablation Study

Ablation MiniF2F-Test Absolute Drop
Full System73.77%
w/o Causal-Pair Expansion (CPE)69.20%4.57%
w/o CPE + Upstream Causal Backtracking / Memory63.11%10.66%
w/o CPE + UCB/Mem + Projected-Candidate Repair57.38%16.39%
w/o CPE + UCB/Mem + PCR + Local FIP Revision45.49%28.28%

Analysis

FIP-states Become More Projection-Ready

We fix the downstream formalizer and vary only the informal source. Final FIP-states from FIPO-Prover reach 53.69% with Gemini 3.1 Pro and 61.48% after lightweight postprocessing, outperforming raw LLM informal proofs, DSP, and LEGO-Prover. This confirms that FIP optimization itself improves formalization utility, not just sampling budget.

Revisions Primarily Explicitize

Analysis of successful trajectories shows that Explicitization dominates the revision distribution (55.2%), while Direction Change accounts for only 6.6%. FIPO-Prover mainly decomposes coarse single-step reasoning into finer, locally checkable transitions, making FIP-to-FP projection easier.

Edit type distribution

FIP transformation grouped by proof-strategy revision categories.

Mechanism vs error type

Recovery-transition usage conditioned on verifier error types.

Transition necessity

Theorem-level necessity of FIP graph transitions in recovered successes.

Executable steps gain

Structural changes in non-direction-changing FIP revisions.

BibTeX

@inproceedings{ma2026fipo,
  author    = {Ma, Jingkun and Wang, Yuchao and Huo, Yujia and Wong, Derek F.},
  title     = {FIPO-Prover: Formalization-Oriented Informal Proof Optimization for Efficient Formal Theorem Proving},
  booktitle = {Proceedings of the 3rd AI for Math Workshop at the 43rd International Conference on Machine Learning (ICML)},
  year      = {2026},
}