pith. sign in
Pith Number

pith:P2FNYCMT

pith:2026:P2FNYCMTM543BVY6F52Z4IVN4S
not attested not anchored not stored refs resolved

MathAtlas: A Benchmark for Autoformalization in the Wild

Davit Babayan, Hafsah Mahmood, Jeffrey Flanigan, Laurel Willey, Liam McCarty, Nilay Patel, Noah Arias, Soli Munoz, Timothy Libman, Victoria Cochran

MathAtlas extracts 52k graduate-level math statements from textbooks to benchmark autoformalization systems.

arxiv:2605.14061 v1 · 2026-05-13 · cs.AI · cs.LG

Add to your LaTeX paper
\usepackage{pith}
\pithnumber{P2FNYCMTM543BVY6F52Z4IVN4S}

Prints a linked badge after your title and injects PDF metadata. Compiles on arXiv. Learn more · Embed verified badge

Record completeness

1 Bitcoin timestamp
2 Internet Archive
3 Author claim open · sign in to claim
4 Citations open
5 Replications open
Portable graph bundle live · download bundle · merged state
The bundle contains the canonical record plus signed events. A mirror can host it anywhere and recompute the same current state with the deterministic merge algorithm.

Claims

C1strongest claim

MathAtlas is high quality but extremely challenging: strong baselines achieve at most 9.8% correctness on theorem statements and 16.7% on definitions. On MA-Hard, the best model achieves only 2.6% correctness.

C2weakest assumption

That the automatic extraction process from textbooks produces accurate, correctly scoped graduate-level statements and that the constructed dependency graph faithfully reflects mathematical prerequisites without introducing extraction errors.

C3one line summary

MathAtlas is the first large-scale benchmark for autoformalizing graduate mathematics, where even strong models reach only 9.8% correctness on theorem statements and drop to 2.6% on the hardest dependency-deep subset.

References

54 extracted · 54 resolved · 3 Pith anchors

[1] The Lean Mathematical Library · doi:10.1145/3372885.3373827
[2] A. Agrawal, S. Gadgil, N. Goyal, A. Narayanan, and A. Tadipatri. Towards a Mathematics Formalisation Assistant using Large Language Models. URL http://arxiv.org/abs/2211. 07524
[3] Z. Azerbayev, B. Piotrowski, and J. Avigad. ProofNet: A Benchmark for Autoformalizing and Formally Proving Undergraduate-Level Mathematics Problems
[4] Z. Azerbayev, H. Schoelkopf, K. Paster, M. D. Santos, S. McAleer, A. Q. Jiang, J. Deng, S. Biderman, and S. Welleck. Llemma: An Open Language Model For Mathematics, Oct. 2023 2023
[5] L. Blecher, G. Cucurull, T. Scialom, and R. Stojnic. Nougat: Neural optical understanding for academic documents, 2023 2023

Formal links

2 machine-checked theorem links

Receipt and verification
First computed 2026-05-17T23:39:12.526508Z
Builder pith-number-builder-2026-05-17-v1
Signature Pith Ed25519 (pith-v1-2026-05) · public key
Schema pith-number/v1.0

Canonical hash

7e8adc09936779b0d71e2f759e22ade4b0af713f39017259c58e9a7bc13daa4d

Aliases

arxiv: 2605.14061 · arxiv_version: 2605.14061v1 · doi: 10.48550/arxiv.2605.14061 · pith_short_12: P2FNYCMTM543 · pith_short_16: P2FNYCMTM543BVY6 · pith_short_8: P2FNYCMT
Agent API
Verify this Pith Number yourself
curl -sH 'Accept: application/ld+json' https://pith.science/pith/P2FNYCMTM543BVY6F52Z4IVN4S \
  | jq -c '.canonical_record' \
  | python3 -c "import sys,json,hashlib; b=json.dumps(json.loads(sys.stdin.read()), sort_keys=True, separators=(',',':'), ensure_ascii=False).encode(); print(hashlib.sha256(b).hexdigest())"
# expect: 7e8adc09936779b0d71e2f759e22ade4b0af713f39017259c58e9a7bc13daa4d
Canonical record JSON
{
  "metadata": {
    "abstract_canon_sha256": "74d2fc61ad397429c54a53a3c8c953b1ddb337746c9c47b077e90ee7cc72012b",
    "cross_cats_sorted": [
      "cs.LG"
    ],
    "license": "http://creativecommons.org/licenses/by/4.0/",
    "primary_cat": "cs.AI",
    "submitted_at": "2026-05-13T19:35:46Z",
    "title_canon_sha256": "548d1dde5b807c3b1ea3382ef7b1edd6616e5abd2f965726c9b4f53db2a60df1"
  },
  "schema_version": "1.0",
  "source": {
    "id": "2605.14061",
    "kind": "arxiv",
    "version": 1
  }
}