Datasets:
Download components/google-math/README.md from MMLA-ORG/Cayley-Math: direct link, hf CLI and curl.
- Browser
- Download file 21.2 kB
-
https://huggingface.co/datasets/MMLA-ORG/Cayley-Math/resolve/refs%2Fpr%2F8/components/google-math/README.md
- Command line
-
hf download hf://datasets/MMLA-ORG/Cayley-Math@refs/pr/8/components/google-math/README.md
-
curl -L -o README.md https://huggingface.co/datasets/MMLA-ORG/Cayley-Math/resolve/refs%2Fpr%2F8/components/google-math/README.md
Google math papers contribution to Cayley-Math
An independently versioned component holding source-derived text from eight Google and Google DeepMind
publications about mathematics, formal mathematics and mathematical reasoning. It follows Cayley-Math's
components/ convention and reuses the record schema, Parquet encoding, split policy and packed-array
format of components/openai-math/. The frozen Phase 1 mirror is unchanged.
What this is, and what it is not
These are artifacts of specific Google math papers (Lean libraries, geometry problem files, generator output, paper notebooks and proof PDFs), not a survey of Google's mathematics output and not a paper-complete reproduction of any one of them. The corpus is a curated projection: a bounded, seeded sample for the generator source, a verbatim text projection for the file sources.
Nothing here establishes mathematical truth. No Lean project was compiled, no proof was checked, no
lake build or Comparator run was performed, and no mathematical verification of any kind took place.
record_type labels the artifact kind, not the validity of its content. Records that contain sorry
placeholders remain placeholders.
The mathematics-dataset subset is output of the Saxton/Grefenstette/Hill/Kohli ICLR 2018 generator at
school-level difficulty. Those pairs are not research mathematics, and the subset is a sampled
projection of the paper's curriculum rather than the paper's dataset. Every record carries
content_origin so consumers can separate verbatim source text from generator_output.
This corpus is mostly seeded generator output, not research mathematics. Measured over the delivered
data, generated_qa_pair records are 477,276 of 480,584 records, which is 99.31% of records (0.993117)
and 35,710,731 of 78,364,888 text characters (45.57%). The formal subset is a small minority by record
count. Lean records total 3,027, namely 535 formal_lean_proof plus 2,492 formal_lean_statement, which
is 0.63% of records (0.006299). Geometry problems, prose proofs and notebooks add 281 more records, so
every non-generator record together is 3,308 records, 0.69% of the corpus (0.006883). Lean text does carry
54.22% of the characters (0.542177) because Lean files are long, so a character-weighted view inverts the
balance while the record-weighted view does not. Read the composition from records, not from bytes.
No Lean compilation and no proof verification was performed on any subset, so no record in this corpus
has a checked proof or a checked statement.
Included sources
Each source is pinned to a commit or dataset revision, and each ships math text under Apache-2.0 software terms with CC-BY 4.0 for non-software materials where the upstream states so.
| Source | Upstream at pinned revision | Records | Groups | Text characters | Record types |
|---|---|---|---|---|---|
alphaproof-nexus |
alphaproof-nexus-results 0647711a7118 |
83 | 51 | 2,229,771 | formal_lean_proof, natural_language_proof |
formal-conjectures |
formal-conjectures 6751c0f04398 |
2,135 | 2,135 | 8,133,507 | formal_lean_statement |
formal-imo |
formal-imo a85d84b97c8e |
261 | 261 | 420,948 | formal_lean_statement |
formal-putnam-like |
formal-putnam-like 636c2d46b3a7 |
96 | 96 | 117,201 | formal_lean_statement |
cryptids |
cryptids 3647718b5df2 |
464 | 464 | 31,638,592 | formal_lean_proof |
mathematics-dataset |
mathematics_dataset 427f45075f84 |
477,276 | 127 | 35,710,731 | generated_qa_pair |
alphageometry |
alphageometry 6777cb586cbb |
267 | 267 | 70,907 | formal_geometry_problem, geometry_rule_table |
mathematics-conjectures |
mathematics_conjectures 8302f2910101 |
2 | 2 | 43,231 | research_notebook |
alphaproof-nexus keeps all 83 records of the earlier standalone build unchanged in content: 71 Lean
proof files plus 12 pypdf-extracted prose proof PDFs over 51 problem groups. Only its split assignment
changed, because splits are recomputed over the merged group set of all eight sources rather than over
its own 51 groups. formal-conjectures is the 1,861-file FormalConjectures/ statement library plus
its 274 helper-library files. formal-imo and formal-putnam-like are Lean 4 problem statements.
cryptids is the whole Lean library, including the Busy Beaver holdout proofs, whose individual files
reach 1.4 MB of mechanically generated state traces. alphageometry contributes 265 geometry problems
in the AlphaGeometry language: 231 JGEX problems, 30 olympiad problems and 4 library examples, plus the
defs.txt and rules.txt rule tables. The two mathematics-conjectures notebooks accompany the Nature
2021 knot-theory and representation-theory paper.
Excluded candidates and why
The widened scan is recorded in full in sources.json (excluded_candidates) and audit/source-scan.json.
The short version:
google-deepmind/eval_hub(putnam_like and live_math) is excluded because every problem carries aCANARY_IDbenchmark-contamination canary and live_math is an actively maintained benchmark whose multimodal problems depend on images absent from this text corpus.deepmind/aqua_rataccompanies a program-synthesis paper, not a mathematics paper.deepmind/math_datasetis a 40 GB pre-generated mirror of the generator already used here.google-research/math-opt-benchmark,google/mathsteps,google-deepmind/alphatensorand every math-named software library (tree-math,mathfu,vector_math.dart,opengrm) ship code and no math text.google-deepmind/cartesian-framesformalises an alignment blog sequence in HOL4 SML, not a mathematics paper.google-deepmind/miniF2Fis a fork of a non-Google corpus.google-deepmind/alphageometry2embeds 26 problems in Python dict literals, and AG1 already ships the same olympiad problems verbatim in the AlphaGeometry language.google/feabench,google/phasecoderandgoogle-research/pointditare physics, audio and vision papers whose titles happen to contain mathematical words.
Files and schema
data/{train,dev,test}.jsonl.gz: source records with required fields text, group, license; unique id, paper_id, split, record_type, source_url, andsource_idnaming the upstream.data/{train,dev,test}.parquet: identical records. Core fields (id,text,group,license,paper_id,split,record_type,source_url) are string columns; all remaining JSON fields are serialised inmetadata_json. Mergejson.loads(metadata_json)with the core columns to reconstruct each JSONL record.sources.json: every included source with its pinned revision, license and measured counts, plus every excluded candidate with its reason.licenses/<source_id>/LICENSE: one upstream license per source;LICENSEat the component root is the Apache-2.0 text shared by all eight.audit/source-files.jsonl: path, sha256, bytes, source_id, kind and pinned source_url for every blob and for the generated corpus file.audit/group-splits.json: the group assignment, the policy, and per-source group counts.audit/extraction.jsonl: per-PDF and per-notebook extraction metrics and quality flags.audit/mathematics-dataset-generation.json: generator commit, seed, per-regime quotas, module list, and measured yield.audit/excluded-duplicates.jsonl,audit/packing.json,audit/source-scan.json.audit/validation-evidence.json: the independent decode and round-trip evidence described below.audit/publish-check.json: the second publish gate, measured from the delivered files after the build.statistics.json: measured records, groups, splits, characters and packed counts, per source.manifest.json: SHA256 and byte size for every delivered file except itself and validation.json.packed/: mmla-type2-packed-v1 arrays, manifest, row attribution and per-record token counts.tools/: the generator driver, the build script and the validation scripts actually run.
Loading
import gzip, json
from pathlib import Path
component = Path('components/google-math')
with gzip.open(component / 'data/train.jsonl.gz', 'rt', encoding='utf-8') as f:
first_record = json.loads(next(f))
import pyarrow.parquet as pq
row = pq.read_table(component / 'data/train.parquet').slice(0, 1).to_pylist()[0]
metadata = json.loads(row.pop('metadata_json'))
first_parquet_record = {**row, **metadata}
Packed training arrays
packed/manifest.json uses mmla-type2-packed-v1: input_ids, labels and role_ids are int32 arrays of
shape [rows, 2048], pad=0, role=0, ignored labels=-100. Each record is encoded as
BOS + SentencePiece(text) + EOS with the included Cayley-Math tokenizer (32768 pieces); window stride is
2048 with one-token lookahead, labels predict the following token, and targets that cross into a new
record's BOS or into padding are masked. Split-specific rows.jsonl maps every non-padding token back to
its record; records.jsonl stores token counts and hashes. This component does not reproduce Phase 1
quotas, packing or splits.
Measured packed counts: train 17691 rows / 36230380 tokens / 35751064 loss-bearing labels; dev 51 / 102872 / 101639; test 91 / 185977 / 185942.
Splits
Splits are assigned per group, never per record, so companions and families stay together: a Lean proof
and its prose proof, a conjecture file and its helper library, or one generated question family. Every
group key is namespaced by its source_id. Assignment sorts SHA256(group) as hex, reserves the first
ceil(1%) of groups for dev, the next ceil(1%) for test, and the rest for train.
audit/group-splits.json freezes every assignment: train
479316 records over 3333 groups,
dev 1233 over 35,
test 35 over 35.
Text projection
Lean, geometry-problem and rule files are byte-identical to the pinned blob, license headers included.
Notebook records join the cell sources of each notebook with a newline; no cell text is altered.
natural_language_proof records come from pypdf page-text extraction and are lossy: math symbols
arrive as Unicode substitutions, ligature artifacts appear, and inline formulas lose their two-dimensional
layout. C0/C1 control characters were replaced and blank-line runs collapsed. Each PDF record carries
extraction_quality = lossy_pdf_text_layer with page and character counts, and audit/extraction.jsonl
holds the same per-file audit.
Verification evidence
tools/validate_google_math.py checks file hashes, source coverage, unique record IDs and unique text,
group split separation, lossless Parquet reconstruction, packed array shapes, causal shift, padding and
boundary masks, and re-encodes every record's token stream against data/*.jsonl.gz. It writes
validation.json; the last run passed all 13 checks
(all_delivered_file_hashes_and_sizes, pinned_sources_metadata_and_excluded_candidates,
source_file_inventory_local_bytes, group_split_policy_recomputed_from_scratch,
lossless_jsonl_parquet_exact_text_uniqueness_and_record_provenance, group_split_separation,
verbatim_lean_source_equality_local_bytes, lean_and_generated_records_survive,
per_file_extraction_audit, dedup_audit_present, generator_audit_seed_and_quota,
all_packed_tokens_reencoded_source_shift_padding_masks_and_attribution, statistics_match_artifacts).
tools/verify_google_math.py is the independent check against the delivered files, and its output is
audit/validation-evidence.json. It decoded 96 packed rows, covering
68 of the 70 train shards plus the single dev and test shards, and for each row (a) confirmed every row
slice equals the re-encoding of the corresponding span of [BOS] + SentencePiece(text) + [EOS] for its
attributed source record and (b) confirmed the decoded text equals the decoded source text modulo
whitespace. It recorded 8 literal
unified diffs of a decoded row against its source record; every diff had 0 lines, whitespace identical.
It round-tripped 3 Parquet rows (indices 0, 239,658 and 479,315) back to their JSONL records with exact
match, and confirmed a formal_lean_proof, a formal_lean_statement, a generated_qa_pair, a
formal_geometry_problem, a natural_language_proof and a research_notebook record all survive with
matching text_sha256. Split separation was recomputed from the data rather than trusted: 3,403 groups
in the audit and 3,403 in the data, no group spanning two splits, train/dev/test at 3,333/35/35 groups.
audit/publish-check.json is a second publish gate with its own code rather
than this component's tools. It agrees with every total above (480,584 records, 3,403 groups, 78,364,888
characters, and the packed counts for all three splits), decodes 12 packed rows across all three splits
and compares 263 attributed segments. All 263 row slices matched their canonical token slices by exact
identity, and all 248 segments that cover a whole record decoded to text identical to the source record
under whitespace normalisation and re-encoded to the identical token sequence. The other 15 segments are
row-boundary fragments that begin or end inside a SentencePiece piece, so they cannot re-encode
fragment-by-fragment; each one's decoded text was still located verbatim inside its attributed record.
It matched 5 of 5 Parquet round-trips exactly and recomputed group separation from data/*.jsonl.gz
directly, finding 0 groups spanning two splits.
The record mix is deliberately stated because it is skewed: generated_qa_pair records are 99.31% of
records but 45.57% of characters, while formal Lean records are 0.63% of records and 54.22% of characters,
and all non-generator records together are 0.69% of records. A consumer wanting formal or prose
mathematics should filter on record_type rather than read the corpus whole.
Publication
The component is delivered as Hugging Face dataset pull request
MMLA-ORG/Cayley-Math discussion #3
on ref refs/pr/3. It carries the whole component plus the top-level README.md config and Collections
update, and it has not been merged into main. Consult that discussion for the current merge state.
A token with write scope or org-admin rights is required to merge it. The token used for this build can
read the repository and open pull requests, but a direct commit to main is rejected with
403 Forbidden: You have read access but not the required permissions for this operation. Merging the
pull request, or re-running hf upload without --create-pr once such a token is available, completes
the submission.
Limitations
- No Lean compilation, proof checking, Comparator run or mathematical verification of any kind.
- The build ran in a sandbox-free subprocess only for the generator;
cryptids,formal-conjectures,formal-imo,formal-putnam-like,alphageometry,alphaproof-nexusand the notebooks were never built or executed, only read as text. - PDF prose proofs are noisy pypdf text; symbol and ligature fidelity is not guaranteed.
- The generator subset is a seeded sample of one curriculum, not the paper's full dataset, and its
questions are school-level rather than research mathematics. One module,
algebra__linear_1d, stopped at 3,055 of its 8,000 quota when upstream's ownassert entropy == 0fired; nothing was repaired, and the shortfall is recorded inaudit/mathematics-dataset-generation.json. formal-conjecturesupstream states that some content derives from third-party sources (Wikipedia, MathOverflow and OEIS under CC-BY-SA 4.0). Per-file reference URLs are retained inside the Lean text; no separate per-file license audit was performed.- Near-duplicate detection, Phase 1 overlap analysis and editorial review were not performed. 1,629
exact duplicate texts were dropped and are listed in
audit/excluded-duplicates.jsonl. - The SentencePiece model normalises runs of spaces, so packed text decodes whitespace-normalised rather than byte-identically to the sources.
- Only problems that upstream published are present, so several subsets are selection-biased.
cryptidscontributes 31.6M of the 78.4M characters because a handful of Busy Beaver traces are enormous, so the corpus is dominated by machine-generated proof script rather than mathematics prose.
Statistics
{
"component": "google-math",
"sources": 8,
"source_commits": {
"alphaproof-nexus": "0647711a71183c1ea492ad60860776617ce1ea88",
"formal-conjectures": "6751c0f043985ac6647db28a45a63afdf21acffe",
"formal-imo": "a85d84b97c8e88a352c136b1e29c355139609bea",
"formal-putnam-like": "636c2d46b3a7d7e77a9f58209b0c17816f8d172b",
"cryptids": "3647718b5df22e1a61462dec0376bd204526285a",
"mathematics-dataset": "427f45075f84b8b9774950196ad63867ca20ffb3",
"alphageometry": "6777cb586cbb46beed28db12dc72c69770b68337",
"mathematics-conjectures": "8302f29101019434e5a0a25807a4e42cd9d96414"
},
"records_before_deduplication": 482213,
"records": 480584,
"excluded_exact_duplicates": 1629,
"groups": 3403,
"record_types": {
"geometry_rule_table": 2,
"formal_geometry_problem": 265,
"formal_lean_proof": 535,
"natural_language_proof": 12,
"formal_lean_statement": 2492,
"research_notebook": 2,
"generated_qa_pair": 477276
},
"text_characters": 78364888,
"per_source": {
"alphaproof-nexus": {
"records": 83,
"groups": 51,
"record_types": {
"formal_lean_proof": 71,
"natural_language_proof": 12
},
"text_characters": 2229771,
"upstream_files": 83,
"split_records": {
"train": 83,
"dev": 0,
"test": 0
}
},
"formal-conjectures": {
"records": 2135,
"groups": 2135,
"record_types": {
"formal_lean_statement": 2135
},
"text_characters": 8133507,
"upstream_files": 2135,
"split_records": {
"train": 2088,
"dev": 23,
"test": 24
}
},
"formal-imo": {
"records": 261,
"groups": 261,
"record_types": {
"formal_lean_statement": 261
},
"text_characters": 420948,
"upstream_files": 261,
"split_records": {
"train": 252,
"dev": 6,
"test": 3
}
},
"formal-putnam-like": {
"records": 96,
"groups": 96,
"record_types": {
"formal_lean_statement": 96
},
"text_characters": 117201,
"upstream_files": 96,
"split_records": {
"train": 93,
"dev": 1,
"test": 2
}
},
"cryptids": {
"records": 464,
"groups": 464,
"record_types": {
"formal_lean_proof": 464
},
"text_characters": 31638592,
"upstream_files": 464,
"split_records": {
"train": 457,
"dev": 2,
"test": 5
}
},
"mathematics-dataset": {
"records": 477276,
"groups": 127,
"record_types": {
"generated_qa_pair": 477276
},
"text_characters": 35710731,
"upstream_files": 1,
"split_records": {
"train": 476076,
"dev": 1200,
"test": 0
}
},
"alphageometry": {
"records": 267,
"groups": 267,
"record_types": {
"geometry_rule_table": 2,
"formal_geometry_problem": 265
},
"text_characters": 70907,
"upstream_files": 5,
"split_records": {
"train": 265,
"dev": 1,
"test": 1
}
},
"mathematics-conjectures": {
"records": 2,
"groups": 2,
"record_types": {
"research_notebook": 2
},
"text_characters": 43231,
"upstream_files": 2,
"split_records": {
"train": 2,
"dev": 0,
"test": 0
}
}
},
"splits": {
"train": {
"records": 479316,
"groups": 3333,
"sources": 8,
"text_characters": 77705566
},
"dev": {
"records": 1233,
"groups": 35,
"sources": 6,
"text_characters": 233131
},
"test": {
"records": 35,
"groups": 35,
"sources": 5,
"text_characters": 426191
}
},
"packed": {
"train": {
"rows": 17691,
"tokens": 36230380,
"loss_bearing_labels": 35751064
},
"dev": {
"rows": 51,
"tokens": 102872,
"loss_bearing_labels": 101639
},
"test": {
"rows": 91,
"tokens": 185977,
"loss_bearing_labels": 185942
}
},
"mathematical_verification": "not_performed",
"lean_compilation": "not_performed"
}