AIProver

An Extensible Neurosymbolic Prover for Mathematical Discovery

A Prover by theorists, for theorists (mathematicians, statisticians, theoretical computer scientists, theoretical physicists, …)

A neuro-symbolic framework pairing auto-formalizing LLMs with SAT/SMT solvers, computer algebra systems, and Lean 4 for proof synthesis, mathematical assistance, and verified discovery.

Auto-formalization Proof synthesis Math assistant Verified discovery
Read the research See publications

A collaboration across


Why AIProver

Bridging informal reasoning and formal verification

A unified pipeline that turns natural-language mathematics into Lean-verified proofs, and acts as a research assistant alongside working mathematicians.

Neuro-symbolic core

An auto-formalizing LLM proposes; SAT/SMT solvers, computer algebra systems, and Lean 4 verify. Every step is grounded in formal logic.

Math assistant

A conversational research aide for mathematicians: answers questions, suggests lemmas, sketches proof strategies, and grounds suggestions in verified knowledge.

Semantic retrieval

Semantic Alignment Models (SAMs) align informal and formal proofs in a shared embedding space, powering RAG context, concept search, and hierarchical proof refinement.

Open & reproducible

Tools, libraries, and benchmarks are released openly so the community can verify, extend, and build on every result.

The Ecosystem

One framework, specialized solvers

AIProver orchestrates domain-specific tools, each targeting a different mathematical area, alongside the Sequencelib formalization project.

GS G-SimpleGroup theory & abstract algebra SL SequencelibLean 4 formalization of the OEIS OF O-ForgeAsymptotic analysis & inequalities QF AI-QFTQuantified formula reasoning