Major Breakthroughs in Lean 4-Based Auto-Formalized Mathematics

January 2025 – January 2026

Last Updated: January 16, 2026 Scope: Autoformalization and formal theorem proving systems utilizing Lean 4


Table of Contents

  1. AlphaProof (DeepMind)

  2. Gauss AI (Math Inc.)

  3. Seed-Prover (ByteDance)

  4. Aristotle (Harmonic)

  5. DeepSeek-Prover-V2 (DeepSeek AI)

  6. Goedel-Prover-V2 (Princeton)

  7. Kimina-Prover (Moonshot AI)

  8. BFS-Prover (ByteDance Seed)

  9. HunyuanProver (Tencent)

  10. InternLM2.5-StepProver (Shanghai AI Lab)

  11. FormaRL (Tsinghua University)

  12. Summary Table


1. AlphaProof (DeepMind)

Achievement: Silver Medal at IMO 2024 | Nature Publication November 2025

AlphaProof is an AlphaZero-inspired agent that learns to find formal proofs through reinforcement learning by training on millions of auto-formalized problems. For the most difficult problems, it uses Test-Time RL, a method of generating and learning from millions of related problem variants at inference time to enable deep, problem-specific adaptation.

Key Results

Technical Approach

Primary Sources


2. Gauss AI (Math Inc.)

Achievement: Strong Prime Number Theorem Formalization | September 2025

Using Gauss, Math Inc. completed a challenge set by Fields Medallist Terence Tao and Alex Kontorovich in January 2024 to formalize the strong Prime Number Theorem (PNT) in Lean. This formalization, which had stalled after 18+ months of human expert effort, was completed in just three weeks.

Key Results

Technical Approach

Significance

"The translation of human mathematics into verifiable machine code has long been a grand challenge. However, the cost of doing so is prohibitive, requiring scarce human expertise."

This represents the first major demonstration of AI dramatically compressing formalization timelines for research-level mathematics.

Primary Sources


3. Seed-Prover (ByteDance)

Achievement: Gold Medal at IMO 2025 | State-of-the-Art Benchmarks

Seed-Prover is a lemma-style whole-proof reasoning model that can iteratively refine proofs based on Lean feedback, proved lemmas, and self-summarization. It achieved gold-medal performance at IMO 2025 with formally verified solutions.

Key Results

Seed-Prover 1.5 Improvements

Technical Approach

Primary Sources


4. Aristotle (Harmonic)

Achievement: Gold Medal at IMO 2025 with Formal Verification | October 2025

Aristotle is an AI system that combines formal verification with informal reasoning, achieving gold-medal-equivalent performance on the 2025 IMO problems. Uniquely, all solutions were automatically verified in Lean 4—no human checking required.

Key Results

Architecture Components

  1. Lean Proof Search System — Monte Carlo Graph Search (MCGS) over Lean 4 proof states

  2. Informal Reasoning System — Generates and formalizes lemmas (autoformalization)

  3. Dedicated Geometry Solver (Yuclid) — Specialized for IMO geometry problems

Technical Innovations

Significance

"When our system outputs the proof, nobody has to look at it. You know it's correct." — Tudor Achim, CEO

This addresses the verification problem: as AI generates more mathematical content, human verification becomes a bottleneck.

Primary Sources


5. DeepSeek-Prover-V2 (DeepSeek AI)

Achievement: State-of-the-Art Neural Theorem Proving | April 2025

DeepSeek-Prover-V2 is an open-source large language model designed for formal theorem proving in Lean 4, using a recursive theorem proving pipeline powered by DeepSeek-V3.

Key Results

Technical Approach

New Benchmark: ProverBench

Primary Sources


6. Goedel-Prover-V2 (Princeton)

Achievement: State-of-the-Art Open-Source Theorem Prover | August 2025

Goedel-Prover-V2 is a series of open-source language models from Princeton University that set a new state-of-the-art in automated theorem proving for Lean 4. The system achieves remarkable results with models 80× smaller than competing systems.

Key Results

Technical Approach

Three key innovations:

  1. Scaffolded Data Synthesis — Generates synthetic tasks of increasing difficulty using Lean's extract_goal to mine failing subgoals from incorrect proofs

  2. Verifier-Guided Self-Correction — Iteratively revises proofs using Lean compiler feedback through long chain-of-thought prompts

  3. Model Averaging — Merges model checkpoints to maintain output diversity in later training stages

Model Variants

ModelMiniF2F pass@32PutnamBench
Goedel-Prover-V2-8B84.6%
Goedel-Prover-V2-32B88.1% (90.4% w/ self-correction)86 problems

Primary Sources


7. Kimina-Prover (Moonshot AI)

Achievement: First Large Formal Reasoning Model with Scaling Laws | April 2025 (72B: July 2025)

Kimina-Prover pioneers a novel reasoning-driven exploration paradigm for formal theorem proving in Lean 4. It was the first system to demonstrate clear performance scaling with model size in neural theorem provers—a property previously unobserved in formal mathematics.

Key Results

Technical Innovations

Scaling Properties

First demonstration that larger LLMs yield stronger formal reasoning when equipped with appropriate RL pipelines:

Model SizeMiniF2F pass@32
1.5B (distilled)~60%
7B (distilled)~68%
72B84.0%

Primary Sources


8. BFS-Prover (ByteDance Seed)

Achievement: Simple Best-First Search Achieves SOTA | February 2025

BFS-Prover challenges the prevailing perception that complex Monte Carlo Tree Search (MCTS) and value functions are necessary for large-scale automated theorem proving. It demonstrates that simple Best-First Search, with targeted optimizations, can outperform sophisticated approaches.

Key Results

Technical Approach

Three key innovations in the expert iteration framework:

  1. Strategic Data Filtering — Excludes problems solvable via beam search node expansion at each iteration round, focusing training on harder cases

  2. DPO from Compiler Feedback — Applies Direct Preference Optimization using state-tactic pairs automatically annotated with Lean compiler error feedback

  3. Length Normalization — Improves BFS search efficiency by normalizing proof length scores

Significance

"This result proves for the first time that, with a sound optimization strategy, a simple BFS method can outperform mainstream sophisticated search algorithms such as MCTS and value functions."

Primary Sources


9. HunyuanProver (Tencent)

Achievement: Guided Tree Search with Critic Models | December 2024 / March 2025

HunyuanProver is a language model fine-tuned from Hunyuan 7B for interactive automatic theorem proving with Lean 4. It introduced scalable data synthesis and guided tree search algorithms for effective "System 2 thinking."

Key Results

Technical Approach

Two core modules:

  1. Scalable Prover-Data Generator

  1. Guided Tree-Search Algorithms

Key Findings

Primary Sources


10. InternLM2.5-StepProver (Shanghai AI Lab)

Achievement: Critic-Guided Search for Deeper Proofs | October 2024 (ICML AI4Math July 2025)

InternLM2.5-StepProver advances automated theorem proving by introducing a prover-critic framework that captures preference information from search trajectories to guide proof search at runtime, enabling discovery of deeper proofs.

Key Results

Technical Approach

Key Insight

"The tree structure constructed during the search process is naturally suited for process supervision. The search trajectories inherently provide the data needed to train the supervision model."

Primary Sources


11. FormaRL (Tsinghua University)

Achievement: Data-Efficient Autoformalization via RL | COLM 2025

FormaRL is a reinforcement learning framework for autoformalization that requires only a small amount of unlabeled data—addressing the critical data scarcity problem in formal verification.

Key Results

Technical Approach

New Benchmark: uproof

Primary Sources


Summary Table

SystemOrganizationKey AchievementDatePrimary Link
AlphaProofDeepMindSilver IMO 2024, Nature publicationNov 2025Nature
GaussMath Inc.Strong PNT formalization in 3 weeksSept 2025Math Inc.
Seed-ProverByteDanceGold IMO 2025 (formal), 99.6% MiniF2FAug 2025arXiv
AristotleHarmonicGold IMO 2025 (formal, auto-verified)Oct 2025arXiv
DeepSeek-Prover-V2DeepSeek88.9% MiniF2F, 49/658 PutnamApril 2025arXiv
Goedel-Prover-V2Princeton90.4% MiniF2F, 86 PutnamBench (open-source SOTA)Aug 2025arXiv
Kimina-ProverMoonshot AI92.2% MiniF2F, first scaling lawsApr/Jul 2025arXiv
BFS-ProverByteDance72.95% MiniF2F via simple BFSFeb 2025arXiv
HunyuanProverTencent68.4% MiniF2F, guided tree searchMar 2025arXiv
InternLM2.5-StepProverShanghai AI Lab65.9% MiniF2F, critic-guided searchOct 2024arXiv
FormaRLTsinghua4-6× accuracy with 859 unlabeled samplesAug 2025OpenReview

1. IMO as the Premier Benchmark

The International Mathematical Olympiad has emerged as the definitive benchmark for measuring AI mathematical reasoning capabilities. In 2025, multiple independent systems achieved gold-medal equivalent performance with formally verified Lean 4 proofs.

2. Formal Verification Advantage

Systems producing Lean 4 proofs offer a crucial advantage: no human verification required. As Tudor Achim (Harmonic) noted: "When our system outputs the proof, nobody has to look at it. You know it's correct."

3. Reinforcement Learning Dominance

Nearly all major advances use RL with formal verification as the reward signal:

4. Simple Methods Can Win

BFS-Prover demonstrated that simple Best-First Search, with careful optimization, can outperform complex MCTS and value function approaches—challenging assumptions about required algorithmic complexity.

5. Scaling Laws Emerge

Kimina-Prover provided the first evidence that larger models yield stronger formal reasoning capabilities when equipped with appropriate RL pipelines—a scaling property previously unobserved in neural theorem provers.

6. Data Efficiency Improvements

7. Scale of Formalization


Benchmarks Reference

BenchmarkDescriptionProblems
MiniF2FCross-system Olympiad-level mathematics488
PutnamBenchUndergraduate competition (Putnam 1962-2023)644
ProverBenchDeepSeek's benchmark including AIME325
Lean WorkbookLarge-scale auto-formalized problems140K+
uproofUndergraduate-level textbook problems5,273

Tools and Infrastructure