Lean Proof Search — Actor–Critic Checkpoints and Deployment Bundles

This repository hosts versioned checkpoints and matched serving artifacts for Lean theorem proving with a source-aware graph neural network policy and an actor–critic reinforcement-learning pipeline.

The application supports PUCT Monte Carlo tree search and best-first search. Optional Probabilistic Logic Networks (PLN) provide additional search guidance. Candidate proofs must pass the application's replay and Lean kernel-certification process before being reported as certified successes.

Application source: https://github.com/iCog-Labs-Dev/Mathematics-ai-atp/tree/MCTS-AC-branch

Model architecture

The source-aware policy combines:

  • A graph encoder for normalized Lean proof-state expressions.
  • A tactic-prediction head.
  • An explicit argument-count head.
  • A GRU argument decoder.
  • A source-selection head that chooses local context or external library arguments.
  • Local-declaration selection and retrieved-library candidate scoring.
  • A critic that estimates proof-state value for reinforcement learning.

External premises are retrieved through a matched text encoder and lemma index. The policy checkpoint, graph vocabularies, query encoder, lemma vectors, and declaration inventory must be used as a compatible deployment set.

These checkpoints are used through the project's custom inference and search code. They are not standalone Transformers text-generation models.

Training and checkpoint provenance

The supervised initialization supplied for this work used a joint source-aware policy followed by local-selection refinement. Its matched retrieval artifacts come from the epoch-5 retriever and index.

Published RL releases may differ in:

  • Starting checkpoint and selected final checkpoint.
  • Search algorithm: PUCT or best-first.
  • Whether PLN guidance was enabled.
  • Training rounds, search budgets, and evaluation settings.
  • Checkpoint selection: periodic-evaluation best or final saved checkpoint.

Consult each release manifest and accompanying release notes for its identity. The serving configuration records search settings but is not a complete training experiment report.

A checkpoint trained using PUCT can also be searched using best-first, provided the model and runtime interfaces remain compatible. The search algorithm is a runtime configuration rather than a separate neural-network architecture.

Repository organization

main contains the repository overview. Published serving bundles live on versioned branches such as v1.0.0.

Each release uses this directory structure:

deployments/mcts/<version>/
├── release.json
├── README.md
├── artifacts/
│   ├── checkpoint.pt
│   ├── prepared/
│   ├── query/
│   ├── index/
│   ├── lemmas.jsonl
│   └── declarations.json
└── config/
    ├── rl.json
    └── service.json

The mcts directory is the deployment format's fixed namespace. It can contain a release configured for either PUCT or best-first search.

The bundle includes:

Component Purpose
Actor–critic checkpoint Trained policy and critic parameters, with any retained checkpoint metadata
Prepared metadata Matching graph representation and node/tactic vocabularies
Query artifact Exported query encoder, tokenizer, and configuration
Lemma index FAISS index, vectors, lemma IDs/names, and identity manifests
Lemma corpus Matched external-premise records
Declaration inventory Available Lean declarations, exclusions, and environment identity
Serving configuration Container paths and service/search configuration
Release manifest File sizes, SHA-256 checksums, artifact bindings, and recorded provenance

Compiled Lean/Mathlib, the custom REPL, Python dependencies, and the PLN runtime are supplied separately by the compatible application runtime or Docker image.

Download and verify a release

Use the downloader from the compatible application checkout:

python -m maths_ai.gnn_inference.scripts.fetch_mcts_deployment \
  --repo iCogLabs-Dev/Lean-Proof-Search-Checkpoints \
  --revision FULL_HUGGING_FACE_RELEASE_COMMIT \
  --version v1.0.0 \
  --manifest-sha256 RELEASE_JSON_SHA256 \
  --output "$HOME/mcts-deployments/v1.0.0"

Replace the commit, version, and manifest checksum with the identifiers provided by the release publisher. Pin the full immutable Hugging Face commit for reproducible deployment.

The downloader verifies file hashes and model/retrieval bindings. The application also checks Lean environment compatibility at startup.

A private repository requires Hugging Face authentication with read access.

The downloaded output is a symlink to a persistent sibling backing directory. Retain both; deleting the backing directory breaks the deployment.

Serving and evaluation

The application's FastAPI service exposes single-goal proving, dataset evaluation jobs, and health/readiness checks.

Search profiles select PUCT or best-first search and may enable PLN. Proposal generation supports greedy or sampled modes where configured.

Docker deployment mounts the verified bundle's artifacts/ directory at /artifacts and its config/ directory at /config.

See the application deployment guides:

PLN-enabled profiles additionally require the provisioned PeTTa/SWI-Prolog runtime. PLN scores guide search; they do not replace Lean proof checking.

Evaluation and interpretation

No aggregate proof-rate improvement is asserted by this model card.

Meaningful comparisons should report:

  • Evaluation dataset and exact theorem cohort.
  • Base and RL checkpoint identities.
  • Search algorithm and proposal mode.
  • PLN setting.
  • Time, node, depth, simulation, and memory budgets.
  • Random seeds.
  • Kernel-certified successes and operational failures.

Results from different theorem cohorts or search budgets do not establish a controlled comparison. Training solve rates should be distinguished from held-out evaluation results.

Limitations

  • Search is bounded and may fail to find an existing proof.
  • Dataset starting states require sufficient Lean context for reconstruction.
  • Tactic and argument proposals can be invalid or unsupported.
  • Retrieval depends on the supplied corpus and deployed Lean environment.
  • A model prediction or graph-level solved candidate is not itself a certified proof.
  • Certification establishes a proof of the admitted Lean starting statement; it does not independently establish that dataset reconstruction preserved the intended original theorem.
  • Matching runtime artifacts are required; arbitrary REPL or Mathlib replacements may invalidate the deployment binding.
  • The serving bundle is not a complete training-resume archive.
  • Checkpoints may contain optimizer state and other retained training metadata. PyTorch checkpoint loading can execute pickle content; load only trusted files.

Deployment: v1.0.2

Selected best.pt from PUCT+PLN training run 20261108-083020. This is a serving bundle, not a complete training-resume archive.

Identifier Value
Repository iCogLabs-Dev/Lean-Proof-Search-Checkpoints
Version v1.0.2
Artifact commit 264df74f6abb33e19a5467664699ad07eb955d89
Manifest SHA-256 ed08e144022c6faf6a565d8e706371a8539bf406f9a756553e6ac42682033fa5
Application commit 2271bdb346d05b749bce773e55a962a6c810acb7

Browse the pinned release.

Compatible runtime

Use application commit 2271bdb346d05b749bce773e55a962a6c810acb7 and its Dockerfile and locked dependencies. The image supplies Lean, the custom Pantograph REPL/client, Mathlib, and PLN. The REPL source identity and compiled Mathlib must match the declaration inventory.

Docker deployment guide.

Download and verify

From that application checkout, in its installed Python environment:

python -m maths_ai.gnn_inference.scripts.fetch_mcts_deployment \
  --repo iCogLabs-Dev/Lean-Proof-Search-Checkpoints \
  --revision 264df74f6abb33e19a5467664699ad07eb955d89 \
  --version v1.0.2 \
  --manifest-sha256 ed08e144022c6faf6a565d8e706371a8539bf406f9a756553e6ac42682033fa5 \
  --output "$HOME/mcts-deployments/v1.0.2"

Keep both the downloaded directory symlink and its hidden sibling backing directory.

CPU or GPU serving

After building the matching image, mount the downloaded directories:

export MATHS_AI_ARTIFACTS="$HOME/mcts-deployments/v1.0.2/artifacts"
export MATHS_AI_CONFIG="$HOME/mcts-deployments/v1.0.2/config"

For CPU:

MATHS_AI_DEVICE=cpu docker compose up -d mcts-api

Alternatively, on an NVIDIA-capable Docker host:

MATHS_AI_DEVICE=cuda:0 docker compose --profile gpu up -d mcts-api-gpu

Choose one service. Both use the same artifacts. Mount paths must be accessible to the Docker daemon.

Verify /ready and a kernel-certified /prove result after startup. Successful upload does not establish successful Docker or GPU execution.

Deployment release: v1.0.3

This serving release contains the selected best.pt from the 100-round PUCT/no-PLN training run 20261112-143713.

Publisher-reported miniF2F validation result: 63/244 (25.82%), using PUCT with 512 simulations, a 512-node limit, and PLN disabled. This is an evaluation of the selected checkpoint, not an additional training update.

Identifier Value
Artifact repository iCogLabs-Dev/Lean-Proof-Search-Checkpoints
Release version v1.0.3
Immutable artifact commit 1a48c749c9b2ec4b0f20d087998569243198db92
SHA-256 of release.json bafeb8c930ac1d6837e12afc5dc3f06ccfc589bc3efaa9c7605028cdc04e1223
Application Git commit 60e4d2bc854c60abdf2a9d0928e902a1c21f0aef
Custom client version 0.7.0
Custom REPL source commit 46a33e1c4e311cdc03ff1a2baa8eaeb36a69ff40

Browse the pinned release.

Compatible runtime

Use the application commit above and its pinned dependencies and Dockerfile. The runtime must match the release's declaration inventory and environment bindings. Downloading the artifacts does not install the runtime.

Docker deployment instructions.

Download and verify

From the compatible application checkout, with its dependencies installed:

python -m maths_ai.gnn_inference.scripts.fetch_mcts_deployment \
  --repo iCogLabs-Dev/Lean-Proof-Search-Checkpoints \
  --revision 1a48c749c9b2ec4b0f20d087998569243198db92 \
  --version v1.0.3 \
  --manifest-sha256 bafeb8c930ac1d6837e12afc5dc3f06ccfc589bc3efaa9c7605028cdc04e1223 \
  --output "$HOME/mcts-deployments/v1.0.3"

The downloader verifies the complete file hashes. The uploader checked manifest bytes and remote file sizes. Upload completion does not establish that a Docker or GPU deployment has passed a smoke test.

This is a serving bundle, not a complete training-resume archive.

Downloads last month

-

Downloads are not tracked for this model. How to track
Video Preview
loading