Add AlphaProof Nexus component metadata

#2
Files changed (42) hide show
  1. README.md +9 -0
  2. components/google-alphaproof-nexus/LICENSE +202 -0
  3. components/google-alphaproof-nexus/README.md +116 -0
  4. components/google-alphaproof-nexus/audit/excluded-duplicates.jsonl +0 -0
  5. components/google-alphaproof-nexus/audit/extraction.jsonl +12 -0
  6. components/google-alphaproof-nexus/audit/group-splits.json +569 -0
  7. components/google-alphaproof-nexus/audit/packing.json +7 -0
  8. components/google-alphaproof-nexus/audit/source-files.jsonl +92 -0
  9. components/google-alphaproof-nexus/audit/spotcheck.json +382 -0
  10. components/google-alphaproof-nexus/data/dev.jsonl.gz +3 -0
  11. components/google-alphaproof-nexus/data/dev.parquet +3 -0
  12. components/google-alphaproof-nexus/data/test.jsonl.gz +3 -0
  13. components/google-alphaproof-nexus/data/test.parquet +3 -0
  14. components/google-alphaproof-nexus/data/train.jsonl.gz +3 -0
  15. components/google-alphaproof-nexus/data/train.parquet +3 -0
  16. components/google-alphaproof-nexus/manifest.json +205 -0
  17. components/google-alphaproof-nexus/packed/dev-000000-input_ids.npy +3 -0
  18. components/google-alphaproof-nexus/packed/dev-000000-labels.npy +3 -0
  19. components/google-alphaproof-nexus/packed/dev-000000-role_ids.npy +3 -0
  20. components/google-alphaproof-nexus/packed/dev-records.jsonl +1 -0
  21. components/google-alphaproof-nexus/packed/dev-rows.jsonl +2 -0
  22. components/google-alphaproof-nexus/packed/manifest.json +196 -0
  23. components/google-alphaproof-nexus/packed/test-000000-input_ids.npy +3 -0
  24. components/google-alphaproof-nexus/packed/test-000000-labels.npy +3 -0
  25. components/google-alphaproof-nexus/packed/test-000000-role_ids.npy +3 -0
  26. components/google-alphaproof-nexus/packed/test-records.jsonl +2 -0
  27. components/google-alphaproof-nexus/packed/test-rows.jsonl +14 -0
  28. components/google-alphaproof-nexus/packed/train-000000-input_ids.npy +3 -0
  29. components/google-alphaproof-nexus/packed/train-000000-labels.npy +3 -0
  30. components/google-alphaproof-nexus/packed/train-000000-role_ids.npy +3 -0
  31. components/google-alphaproof-nexus/packed/train-000001-input_ids.npy +3 -0
  32. components/google-alphaproof-nexus/packed/train-000001-labels.npy +3 -0
  33. components/google-alphaproof-nexus/packed/train-000001-role_ids.npy +3 -0
  34. components/google-alphaproof-nexus/packed/train-records.jsonl +80 -0
  35. components/google-alphaproof-nexus/packed/train-rows.jsonl +0 -0
  36. components/google-alphaproof-nexus/sources.json +28 -0
  37. components/google-alphaproof-nexus/statistics.json +64 -0
  38. components/google-alphaproof-nexus/tokenizer/tokenizer.model +3 -0
  39. components/google-alphaproof-nexus/tools/build_google_alphaproof_nexus.py +588 -0
  40. components/google-alphaproof-nexus/tools/spotcheck_google_alphaproof_nexus.py +233 -0
  41. components/google-alphaproof-nexus/tools/validate_google_alphaproof_nexus.py +281 -0
  42. components/google-alphaproof-nexus/validation.json +78 -0
README.md CHANGED
@@ -39,6 +39,14 @@ configs:
39
  path: components/openai-math/data/dev.parquet
40
  - split: test
41
  path: components/openai-math/data/test.parquet
 
 
 
 
 
 
 
 
42
  ---
43
 
44
  # Cayley-Math
@@ -49,6 +57,7 @@ Cayley-Math is a collaboratively maintained collection of mathematical text for
49
  | --- | --- | --- |
50
  | Anholon Phase 1 mathematics mirror | [Phase 1 bundle](phase1/README.md) | Anholon data maintainers |
51
  | OpenAI mathematics manuscripts | [OpenAI Math component](components/openai-math/README.md) | Source extraction and OMP DeepSeek metadata |
 
52
  | Contributed mathematics collections | [Component guide](components/README.md) | Each component's contributors |
53
 
54
  ## Phase 1 Files
 
39
  path: components/openai-math/data/dev.parquet
40
  - split: test
41
  path: components/openai-math/data/test.parquet
42
+ - config_name: google-alphaproof-nexus
43
+ data_files:
44
+ - split: train
45
+ path: components/google-alphaproof-nexus/data/train.parquet
46
+ - split: validation
47
+ path: components/google-alphaproof-nexus/data/dev.parquet
48
+ - split: test
49
+ path: components/google-alphaproof-nexus/data/test.parquet
50
  ---
51
 
52
  # Cayley-Math
 
57
  | --- | --- | --- |
58
  | Anholon Phase 1 mathematics mirror | [Phase 1 bundle](phase1/README.md) | Anholon data maintainers |
59
  | OpenAI mathematics manuscripts | [OpenAI Math component](components/openai-math/README.md) | Source extraction and OMP DeepSeek metadata |
60
+ | Google AlphaProof Nexus results | [AlphaProof Nexus component](components/google-alphaproof-nexus/README.md) | Verbatim Lean sources and pypdf prose extraction |
61
  | Contributed mathematics collections | [Component guide](components/README.md) | Each component's contributors |
62
 
63
  ## Phase 1 Files
components/google-alphaproof-nexus/LICENSE ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
176
+
177
+ END OF TERMS AND CONDITIONS
178
+
179
+ APPENDIX: How to apply the Apache License to your work.
180
+
181
+ To apply the Apache License to your work, attach the following
182
+ boilerplate notice, with the fields enclosed by brackets "[]"
183
+ replaced with your own identifying information. (Don't include
184
+ the brackets!) The text should be enclosed in the appropriate
185
+ comment syntax for the file format. We also recommend that a
186
+ file or class name and description of purpose be included on the
187
+ same "printed page" as the copyright notice for easier
188
+ identification within third-party archives.
189
+
190
+ Copyright [yyyy] [name of copyright owner]
191
+
192
+ Licensed under the Apache License, Version 2.0 (the "License");
193
+ you may not use this file except in compliance with the License.
194
+ You may obtain a copy of the License at
195
+
196
+ http://www.apache.org/licenses/LICENSE-2.0
197
+
198
+ Unless required by applicable law or agreed to in writing, software
199
+ distributed under the License is distributed on an "AS IS" BASIS,
200
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201
+ See the License for the specific language governing permissions and
202
+ limitations under the License.
components/google-alphaproof-nexus/README.md ADDED
@@ -0,0 +1,116 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # Google AlphaProof Nexus contribution to Cayley-Math
2
+
3
+ This independently versioned component contains source-derived mathematical text from Google DeepMind's
4
+ AlphaProof Nexus result repository at commit 0647711a71183c1ea492ad60860776617ce1ea88. It follows Cayley-Math's components/ convention and
5
+ reuses the record schema, Parquet encoding, split policy and packed-array format of
6
+ `components/openai-math/`. The frozen Phase 1 mirror is unchanged. Upstream works are attributed to Google
7
+ DeepMind and distributed under Apache-2.0; see LICENSE, sources.json, and audit/.
8
+
9
+ ## What this is, and what it is not
10
+
11
+ These are proofs of **specific solved problems**, not a general corpus or a proof-search benchmark. Upstream
12
+ publishes only the problems where AlphaProof Nexus found a successful proof, so the collection is
13
+ selection-biased and small. Every upstream file carries Google's Apache-2.0 copyright header; the Lean files
14
+ retain it verbatim.
15
+
16
+ No Lean compilation, `lake build`, Comparator run, or mathematical verification was performed. The component
17
+ does not establish mathematical correctness of any statement, formalization, or proof. Upstream states that
18
+ the Lean outputs were mechanically formalized, but this contribution only transports text.
19
+
20
+ ## Coverage
21
+
22
+ The pinned snapshot has **92 files**, including **71 `.lean` files** and
23
+ **12 natural-language proof PDFs**. Projection yields **83 records** across
24
+ **51 problem groups**: **71** `formal_lean_proof`
25
+ and **12** `natural_language_proof`. Exact text
26
+ deduplication removed 0 records; audit/excluded-duplicates.jsonl lists them.
27
+
28
+ Splits are assigned per problem group so a Lean proof and its prose proof never cross train/dev/test.
29
+ Assignment sorts SHA256(group), reserves the first ceil(1%) of groups for development, the next ceil(1%) for
30
+ test, and the rest for training. Assigned counts: train 80 records over
31
+ 49 groups, dev 1 over
32
+ 1, test 2 over
33
+ 1.
34
+
35
+ Grouping derives from the problem stem, for example `erdos_152` from `erdos_152.lean` plus `erdos152.pdf`,
36
+ `oeis_51293`, or `ai_hilbert_functions` from `hilbert_functions_*.lean` plus `hilbert.pdf`. Where no filename
37
+ pairing exists, the containing problem directory is the group; audit/group-splits.json records each
38
+ assignment rule and the unpaired directories. Stacks Project files are flat in one directory and therefore
39
+ share the single group `stacks_project`.
40
+
41
+ ## Text projection
42
+
43
+ `formal_lean_proof` records are the upstream file text, byte-identical to the pinned blob, license header
44
+ included. `natural_language_proof` records come from pypdf page-text extraction of each PDF. Extraction is
45
+ lossy: math symbols arrive as Unicode substitutions, ligature and encoding artifacts appear (for example
46
+ `Erd˝os`), and inline formulas lose their two-dimensional layout. C0/C1 control characters were replaced
47
+ and blank-line runs collapsed; nothing else was altered. Each PDF record carries `extraction_quality =
48
+ lossy_pdf_text_layer` plus page count, character counts, non-ASCII ratio and artifact counts in
49
+ `extraction_metrics`; audit/extraction.jsonl holds the same per-PDF audit. Treat prose records as noisy text,
50
+ not as faithful renderings of the PDFs.
51
+
52
+ ## Files and schema
53
+
54
+ - data/{train,dev,test}.jsonl.gz: source records with required fields text, group, license; unique id, paper_id, split, record_type, source_url, and provenance.
55
+ - 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 serialized in metadata_json. Merge json.loads(metadata_json) with the core columns to reconstruct each JSONL record.
56
+ - audit/source-files.jsonl: path, sha256, bytes and pinned source_url for every upstream blob.
57
+ - audit/group-splits.json: group assignments, per-group rule and unpaired directories.
58
+ - audit/extraction.jsonl: per-PDF extraction metrics and quality flag.
59
+ - audit/excluded-duplicates.jsonl: exact-text duplicates removed before packaging.
60
+ - statistics.json: measured record, group, split and packed-token counts.
61
+ - manifest.json: SHA256 and byte size for delivered files except itself and validation.json.
62
+ - packed/: mmla-type2-packed-v1 arrays, manifest, row attribution and per-record token counts.
63
+ - tools/: the build and validation scripts used to produce and check these files.
64
+
65
+ ## Loading
66
+
67
+ ~~~python
68
+ import gzip, json
69
+ from pathlib import Path
70
+ component = Path('components/google-alphaproof-nexus')
71
+ with gzip.open(component / 'data/train.jsonl.gz', 'rt', encoding='utf-8') as f:
72
+ first_record = json.loads(next(f))
73
+
74
+ import pyarrow.parquet as pq
75
+ row = pq.read_table(component / 'data/train.parquet').slice(0, 1).to_pylist()[0]
76
+ metadata = json.loads(row.pop('metadata_json'))
77
+ first_parquet_record = {**row, **metadata}
78
+ ~~~
79
+
80
+ ## Packed training arrays
81
+
82
+ packed/manifest.json uses `mmla-type2-packed-v1`: input_ids, labels and role_ids are int32 arrays of shape
83
+ [rows, 2048], pad=0, role=0, ignored labels=-100. Each record is encoded as BOS + SentencePiece(text) +
84
+ EOS with the included Cayley-Math tokenizer (32768 pieces); window stride is 2048 with one-token lookahead,
85
+ labels predict the following token, and targets that cross into a new record's BOS or into padding are masked.
86
+ Split-specific rows.jsonl maps every non-padding token back to its record; records.jsonl stores token counts
87
+ and hashes. This component does not reproduce Phase 1 quotas, packing, or splits.
88
+
89
+ Measured packed counts: train 491 rows / 1004469 tokens / 1004389 loss-bearing labels; dev 2 / 3144 / 3143; test 14 / 28411 / 28409.
90
+
91
+ ## Reproduction and verification
92
+
93
+ ~~~sh
94
+ python tools/build_google_alphaproof_nexus.py --component . --source-root /path/to/pinned/alphaproof-nexus-results --tokenizer tokenizer/tokenizer.model --pack
95
+ python tools/validate_google_alphaproof_nexus.py --component . --source-root /path/to/pinned/alphaproof-nexus-results
96
+ python tools/spotcheck_google_alphaproof_nexus.py --component . --write audit/spotcheck.json
97
+ ~~~
98
+
99
+ Dependencies: Python 3.13, numpy, sentencepiece, pyarrow, pypdf. The validator checks file hashes, source
100
+ coverage, exact-text uniqueness, group split separation, lossless Parquet reconstruction, packed array
101
+ shapes, causal shift, padding and boundary masks, token-level re-encoding of every record, and row
102
+ attribution. The spot check decodes packed rows back to text and round-trips a Parquet row to its JSONL
103
+ record. Neither script assesses theorem truth or proof completeness.
104
+
105
+ ## Limitations
106
+
107
+ - PDF extraction is lossy for symbols, ligatures, and formula layout; prose records are noisy.
108
+ - Only solved problems are present upstream, so the collection is not representative.
109
+ - No Lean compilation or proof verification; `lean-toolchain` pins `leanprover/lean4:v4.27.0` and
110
+ `formal_conjectures` is a required dependency, but this snapshot was not built.
111
+ - Near-duplicate detection, Phase 1 overlap analysis, and editorial review were not performed.
112
+ - The SentencePiece model cannot reproduce repeated runs of spaces: decoding drops them, so every packed row
113
+ re-encodes to the same token IDs while its decoded text equals the source only after collapsing whitespace
114
+ runs. This is inherited from the shared Cayley-Math tokenizer and applies to both record types.
115
+ - 12 records are prose and 1 test group (`ai_graph_reconstruction`) carries both record types, so the test
116
+ split holds 2 of the 83 records. Monitoring splits are small by construction at this corpus size.
components/google-alphaproof-nexus/audit/excluded-duplicates.jsonl ADDED
File without changes
components/google-alphaproof-nexus/audit/extraction.jsonl ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"artifact_characters":{"c0_c1_control":146,"combining_mark":6},"characters":7842,"control_characters_replaced":0,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":149,"non_ascii_ratio":0.019,"page_characters":[2123,1881,1457,2377],"pages":4,"raw_characters":7841,"source_path":"NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf"}
2
+ {"artifact_characters":{"c0_c1_control":248,"combining_mark":5},"characters":11242,"control_characters_replaced":6,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":387,"non_ascii_ratio":0.034424,"page_characters":[2123,1251,1516,1713,1540,1510,1585],"pages":7,"raw_characters":11244,"source_path":"NaturalLanguageProofs/AICollaborator/hilbert.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/AICollaborator/hilbert.pdf"}
3
+ {"artifact_characters":{"c0_c1_control":86},"characters":4253,"control_characters_replaced":5,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":93,"non_ascii_ratio":0.021867,"page_characters":[2372,1879],"pages":2,"raw_characters":4252,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos12i.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos12i.pdf"}
4
+ {"artifact_characters":{"c0_c1_control":77},"characters":4186,"control_characters_replaced":11,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":71,"non_ascii_ratio":0.016961,"page_characters":[2664,1522],"pages":2,"raw_characters":4187,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos12ii.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos12ii.pdf"}
5
+ {"artifact_characters":{"c0_c1_control":52},"characters":3493,"control_characters_replaced":0,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":40,"non_ascii_ratio":0.011451,"page_characters":[3211,280],"pages":2,"raw_characters":3492,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos125.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos125.pdf"}
6
+ {"artifact_characters":{"c0_c1_control":22},"characters":1483,"control_characters_replaced":0,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":17,"non_ascii_ratio":0.011463,"page_characters":[1482],"pages":1,"raw_characters":1482,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos138.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos138.pdf"}
7
+ {"artifact_characters":{"c0_c1_control":56},"characters":3266,"control_characters_replaced":3,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":83,"non_ascii_ratio":0.025413,"page_characters":[2368,896],"pages":2,"raw_characters":3265,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos152.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos152.pdf"}
8
+ {"artifact_characters":{"c0_c1_control":77},"characters":3305,"control_characters_replaced":3,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":51,"non_ascii_ratio":0.015431,"page_characters":[2225,1078],"pages":2,"raw_characters":3304,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos26.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos26.pdf"}
9
+ {"artifact_characters":{"c0_c1_control":46},"characters":2695,"control_characters_replaced":6,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":52,"non_ascii_ratio":0.019295,"page_characters":[2338,355],"pages":2,"raw_characters":2694,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos741i.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos741i.pdf"}
10
+ {"artifact_characters":{"c0_c1_control":33},"characters":2181,"control_characters_replaced":3,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":28,"non_ascii_ratio":0.012838,"page_characters":[2180],"pages":1,"raw_characters":2180,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos741ii.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos741ii.pdf"}
11
+ {"artifact_characters":{"c0_c1_control":66},"characters":3950,"control_characters_replaced":0,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":34,"non_ascii_ratio":0.008608,"page_characters":[2204,1744],"pages":2,"raw_characters":3949,"source_path":"NaturalLanguageProofs/ErdosProblems/erdos846.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos846.pdf"}
12
+ {"artifact_characters":{"c0_c1_control":146},"characters":4519,"control_characters_replaced":62,"extraction_quality":"lossy_pdf_text_layer","non_ascii_characters":89,"non_ascii_ratio":0.019695,"page_characters":[1948,1995,595],"pages":3,"raw_characters":4540,"source_path":"NaturalLanguageProofs/OEIS/oeis_51293.pdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/OEIS/oeis_51293.pdf"}
components/google-alphaproof-nexus/audit/group-splits.json ADDED
@@ -0,0 +1,569 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "cayley-group-splits-v1",
3
+ "source_commit": "0647711a71183c1ea492ad60860776617ce1ea88",
4
+ "policy": "sorted SHA256(group), first ceil(1%) groups to dev, next ceil(1%) to test, rest train",
5
+ "splits": {
6
+ "oeis_243106": "dev",
7
+ "ai_graph_reconstruction": "test",
8
+ "oeis_175386": "train",
9
+ "oeis_289411": "train",
10
+ "oeis_341254": "train",
11
+ "stacks_project": "train",
12
+ "oeis_a211417": "train",
13
+ "oeis_194806": "train",
14
+ "oeis_a258667": "train",
15
+ "ai_additivecombinatorics": "train",
16
+ "erdos_846": "train",
17
+ "ai_hilbert_functions": "train",
18
+ "oeis_363347": "train",
19
+ "oeis_a300997": "train",
20
+ "oeis_306424": "train",
21
+ "oeis_a309132": "train",
22
+ "erdos_12": "train",
23
+ "ai_graphs": "train",
24
+ "oeis_323557": "train",
25
+ "oeis_113254": "train",
26
+ "ai_optimization": "train",
27
+ "erdos_741": "train",
28
+ "oeis_307865": "train",
29
+ "oeis_62567": "train",
30
+ "oeis_51293": "train",
31
+ "oeis_a368692": "train",
32
+ "erdos_138": "train",
33
+ "erdos_125": "train",
34
+ "oeis_372761": "train",
35
+ "oeis_228143": "train",
36
+ "oeis_2897": "train",
37
+ "oeis_103311": "train",
38
+ "oeis_256012": "train",
39
+ "oeis_267581": "train",
40
+ "erdos_152": "train",
41
+ "oeis_340737": "train",
42
+ "oeis_a028859": "train",
43
+ "oeis_278070": "train",
44
+ "oeis_a363102": "train",
45
+ "oeis_a091669": "train",
46
+ "oeis_a224515": "train",
47
+ "oeis_271591": "train",
48
+ "oeis_a237271": "train",
49
+ "oeis_108": "train",
50
+ "erdos_26": "train",
51
+ "oeis_a382590": "train",
52
+ "oeis_a325046": "train",
53
+ "oeis_282779": "train",
54
+ "oeis_227582": "train",
55
+ "oeis_248802": "train",
56
+ "ai_quantumoptics": "train"
57
+ },
58
+ "rules": {
59
+ "ai_additivecombinatorics": [
60
+ "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/AdditiveCombinatorics"
61
+ ],
62
+ "ai_graph_reconstruction": [
63
+ "paired with NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf by problem stem"
64
+ ],
65
+ "ai_graphs": [
66
+ "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/Graphs"
67
+ ],
68
+ "ai_hilbert_functions": [
69
+ "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
70
+ ],
71
+ "ai_optimization": [
72
+ "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/Optimization"
73
+ ],
74
+ "ai_quantumoptics": [
75
+ "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/QuantumOptics"
76
+ ],
77
+ "erdos_12": [
78
+ "erdos problem number from the filename stem"
79
+ ],
80
+ "erdos_125": [
81
+ "erdos problem number from the filename stem"
82
+ ],
83
+ "erdos_138": [
84
+ "erdos problem number from the filename stem"
85
+ ],
86
+ "erdos_152": [
87
+ "erdos problem number from the filename stem"
88
+ ],
89
+ "erdos_26": [
90
+ "erdos problem number from the filename stem"
91
+ ],
92
+ "erdos_741": [
93
+ "erdos problem number from the filename stem"
94
+ ],
95
+ "erdos_846": [
96
+ "erdos problem number from the filename stem"
97
+ ],
98
+ "oeis_103311": [
99
+ "OEIS sequence identifier from the filename stem"
100
+ ],
101
+ "oeis_108": [
102
+ "OEIS sequence identifier from the filename stem"
103
+ ],
104
+ "oeis_113254": [
105
+ "OEIS sequence identifier from the filename stem"
106
+ ],
107
+ "oeis_175386": [
108
+ "OEIS sequence identifier from the filename stem"
109
+ ],
110
+ "oeis_194806": [
111
+ "OEIS sequence identifier from the filename stem"
112
+ ],
113
+ "oeis_227582": [
114
+ "OEIS sequence identifier from the filename stem"
115
+ ],
116
+ "oeis_228143": [
117
+ "OEIS sequence identifier from the filename stem"
118
+ ],
119
+ "oeis_243106": [
120
+ "OEIS sequence identifier from the filename stem"
121
+ ],
122
+ "oeis_248802": [
123
+ "OEIS sequence identifier from the filename stem"
124
+ ],
125
+ "oeis_256012": [
126
+ "OEIS sequence identifier from the filename stem"
127
+ ],
128
+ "oeis_267581": [
129
+ "OEIS sequence identifier from the filename stem"
130
+ ],
131
+ "oeis_271591": [
132
+ "OEIS sequence identifier from the filename stem"
133
+ ],
134
+ "oeis_278070": [
135
+ "OEIS sequence identifier from the filename stem"
136
+ ],
137
+ "oeis_282779": [
138
+ "OEIS sequence identifier from the filename stem"
139
+ ],
140
+ "oeis_289411": [
141
+ "OEIS sequence identifier from the filename stem"
142
+ ],
143
+ "oeis_2897": [
144
+ "OEIS sequence identifier from the filename stem"
145
+ ],
146
+ "oeis_306424": [
147
+ "OEIS sequence identifier from the filename stem"
148
+ ],
149
+ "oeis_307865": [
150
+ "OEIS sequence identifier from the filename stem"
151
+ ],
152
+ "oeis_323557": [
153
+ "OEIS sequence identifier from the filename stem"
154
+ ],
155
+ "oeis_340737": [
156
+ "OEIS sequence identifier from the filename stem"
157
+ ],
158
+ "oeis_341254": [
159
+ "OEIS sequence identifier from the filename stem"
160
+ ],
161
+ "oeis_363347": [
162
+ "OEIS sequence identifier from the filename stem"
163
+ ],
164
+ "oeis_372761": [
165
+ "OEIS sequence identifier from the filename stem"
166
+ ],
167
+ "oeis_51293": [
168
+ "OEIS sequence identifier from the filename stem"
169
+ ],
170
+ "oeis_62567": [
171
+ "OEIS sequence identifier from the filename stem"
172
+ ],
173
+ "oeis_a028859": [
174
+ "OEIS sequence identifier from the filename stem"
175
+ ],
176
+ "oeis_a091669": [
177
+ "OEIS sequence identifier from the filename stem"
178
+ ],
179
+ "oeis_a211417": [
180
+ "OEIS sequence identifier from the filename stem"
181
+ ],
182
+ "oeis_a224515": [
183
+ "OEIS sequence identifier from the filename stem"
184
+ ],
185
+ "oeis_a237271": [
186
+ "OEIS sequence identifier from the filename stem"
187
+ ],
188
+ "oeis_a258667": [
189
+ "OEIS sequence identifier from the filename stem"
190
+ ],
191
+ "oeis_a300997": [
192
+ "OEIS sequence identifier from the filename stem"
193
+ ],
194
+ "oeis_a309132": [
195
+ "OEIS sequence identifier from the filename stem"
196
+ ],
197
+ "oeis_a325046": [
198
+ "OEIS sequence identifier from the filename stem"
199
+ ],
200
+ "oeis_a363102": [
201
+ "OEIS sequence identifier from the filename stem"
202
+ ],
203
+ "oeis_a368692": [
204
+ "OEIS sequence identifier from the filename stem"
205
+ ],
206
+ "oeis_a382590": [
207
+ "OEIS sequence identifier from the filename stem"
208
+ ],
209
+ "stacks_project": [
210
+ "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
211
+ ]
212
+ },
213
+ "unpaired_directories": [
214
+ {
215
+ "group": "ai_additivecombinatorics",
216
+ "directory": "APNOutputs/AICollaborator/AdditiveCombinatorics"
217
+ },
218
+ {
219
+ "group": "ai_graphs",
220
+ "directory": "APNOutputs/AICollaborator/Graphs"
221
+ },
222
+ {
223
+ "group": "ai_optimization",
224
+ "directory": "APNOutputs/AICollaborator/Optimization"
225
+ },
226
+ {
227
+ "group": "ai_quantumoptics",
228
+ "directory": "APNOutputs/AICollaborator/QuantumOptics"
229
+ },
230
+ {
231
+ "group": "stacks_project",
232
+ "directory": "APNOutputs/StacksProject"
233
+ }
234
+ ],
235
+ "assignments": {
236
+ "APNOutputs/AICollaborator/AdditiveCombinatorics/57.lean": {
237
+ "group": "ai_additivecombinatorics",
238
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/AdditiveCombinatorics"
239
+ },
240
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_1.lean": {
241
+ "group": "ai_hilbert_functions",
242
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
243
+ },
244
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_2.lean": {
245
+ "group": "ai_hilbert_functions",
246
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
247
+ },
248
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_3.lean": {
249
+ "group": "ai_hilbert_functions",
250
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
251
+ },
252
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_4.lean": {
253
+ "group": "ai_hilbert_functions",
254
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
255
+ },
256
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_5.lean": {
257
+ "group": "ai_hilbert_functions",
258
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
259
+ },
260
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_7.lean": {
261
+ "group": "ai_hilbert_functions",
262
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
263
+ },
264
+ "APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_8.lean": {
265
+ "group": "ai_hilbert_functions",
266
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem"
267
+ },
268
+ "APNOutputs/AICollaborator/Graphs/GraphConjecture2.lean": {
269
+ "group": "ai_graphs",
270
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/Graphs"
271
+ },
272
+ "APNOutputs/AICollaborator/Graphs/bipartite_graph_reconstruction_conjecture_2.lean": {
273
+ "group": "ai_graph_reconstruction",
274
+ "rule": "paired with NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf by problem stem"
275
+ },
276
+ "APNOutputs/AICollaborator/Optimization/LastIterateConvergence.lean": {
277
+ "group": "ai_optimization",
278
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/Optimization"
279
+ },
280
+ "APNOutputs/AICollaborator/QuantumOptics/35.lean": {
281
+ "group": "ai_quantumoptics",
282
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/QuantumOptics"
283
+ },
284
+ "APNOutputs/AICollaborator/QuantumOptics/MonochromaticQuantumGraph.lean": {
285
+ "group": "ai_quantumoptics",
286
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/QuantumOptics"
287
+ },
288
+ "APNOutputs/ErdosProblems/erdos_12.parts.i.lean": {
289
+ "group": "erdos_12",
290
+ "rule": "erdos problem number from the filename stem"
291
+ },
292
+ "APNOutputs/ErdosProblems/erdos_12.parts.ii.lean": {
293
+ "group": "erdos_12",
294
+ "rule": "erdos problem number from the filename stem"
295
+ },
296
+ "APNOutputs/ErdosProblems/erdos_125.variants.positive_lower_density.lean": {
297
+ "group": "erdos_125",
298
+ "rule": "erdos problem number from the filename stem"
299
+ },
300
+ "APNOutputs/ErdosProblems/erdos_138.variants.difference.lean": {
301
+ "group": "erdos_138",
302
+ "rule": "erdos problem number from the filename stem"
303
+ },
304
+ "APNOutputs/ErdosProblems/erdos_152.lean": {
305
+ "group": "erdos_152",
306
+ "rule": "erdos problem number from the filename stem"
307
+ },
308
+ "APNOutputs/ErdosProblems/erdos_26.variants.tenenbaum.lean": {
309
+ "group": "erdos_26",
310
+ "rule": "erdos problem number from the filename stem"
311
+ },
312
+ "APNOutputs/ErdosProblems/erdos_741.parts.i.lean": {
313
+ "group": "erdos_741",
314
+ "rule": "erdos problem number from the filename stem"
315
+ },
316
+ "APNOutputs/ErdosProblems/erdos_741.parts.ii.lean": {
317
+ "group": "erdos_741",
318
+ "rule": "erdos problem number from the filename stem"
319
+ },
320
+ "APNOutputs/ErdosProblems/erdos_846.lean": {
321
+ "group": "erdos_846",
322
+ "rule": "erdos problem number from the filename stem"
323
+ },
324
+ "APNOutputs/OEIS/A224515_conjecture_existence.lean": {
325
+ "group": "oeis_a224515",
326
+ "rule": "OEIS sequence identifier from the filename stem"
327
+ },
328
+ "APNOutputs/OEIS/A309132_conjecture_carmichael.lean": {
329
+ "group": "oeis_a309132",
330
+ "rule": "OEIS sequence identifier from the filename stem"
331
+ },
332
+ "APNOutputs/OEIS/A382590_conjecture_kth_prime_factor_is_eventually_periodic.lean": {
333
+ "group": "oeis_a382590",
334
+ "rule": "OEIS sequence identifier from the filename stem"
335
+ },
336
+ "APNOutputs/OEIS/a091669_conjecture_primitive_root.lean": {
337
+ "group": "oeis_a091669",
338
+ "rule": "OEIS sequence identifier from the filename stem"
339
+ },
340
+ "APNOutputs/OEIS/a325046_odd_terms_at_k_times_k_plus_1.lean": {
341
+ "group": "oeis_a325046",
342
+ "rule": "OEIS sequence identifier from the filename stem"
343
+ },
344
+ "APNOutputs/OEIS/oeis_103311_conjecture_0.lean": {
345
+ "group": "oeis_103311",
346
+ "rule": "OEIS sequence identifier from the filename stem"
347
+ },
348
+ "APNOutputs/OEIS/oeis_108_conjecture_2.lean": {
349
+ "group": "oeis_108",
350
+ "rule": "OEIS sequence identifier from the filename stem"
351
+ },
352
+ "APNOutputs/OEIS/oeis_113254_conjecture_0.lean": {
353
+ "group": "oeis_113254",
354
+ "rule": "OEIS sequence identifier from the filename stem"
355
+ },
356
+ "APNOutputs/OEIS/oeis_175386_conjecture_0.lean": {
357
+ "group": "oeis_175386",
358
+ "rule": "OEIS sequence identifier from the filename stem"
359
+ },
360
+ "APNOutputs/OEIS/oeis_194806_conjecture_0.lean": {
361
+ "group": "oeis_194806",
362
+ "rule": "OEIS sequence identifier from the filename stem"
363
+ },
364
+ "APNOutputs/OEIS/oeis_227582_conjecture_0.lean": {
365
+ "group": "oeis_227582",
366
+ "rule": "OEIS sequence identifier from the filename stem"
367
+ },
368
+ "APNOutputs/OEIS/oeis_228143_conjecture_1.lean": {
369
+ "group": "oeis_228143",
370
+ "rule": "OEIS sequence identifier from the filename stem"
371
+ },
372
+ "APNOutputs/OEIS/oeis_243106_conjecture_0.lean": {
373
+ "group": "oeis_243106",
374
+ "rule": "OEIS sequence identifier from the filename stem"
375
+ },
376
+ "APNOutputs/OEIS/oeis_248802_conjecture_0.lean": {
377
+ "group": "oeis_248802",
378
+ "rule": "OEIS sequence identifier from the filename stem"
379
+ },
380
+ "APNOutputs/OEIS/oeis_248802_conjecture_4.lean": {
381
+ "group": "oeis_248802",
382
+ "rule": "OEIS sequence identifier from the filename stem"
383
+ },
384
+ "APNOutputs/OEIS/oeis_256012_conjecture_0.lean": {
385
+ "group": "oeis_256012",
386
+ "rule": "OEIS sequence identifier from the filename stem"
387
+ },
388
+ "APNOutputs/OEIS/oeis_267581_conjecture_0.lean": {
389
+ "group": "oeis_267581",
390
+ "rule": "OEIS sequence identifier from the filename stem"
391
+ },
392
+ "APNOutputs/OEIS/oeis_271591_conjecture_0.lean": {
393
+ "group": "oeis_271591",
394
+ "rule": "OEIS sequence identifier from the filename stem"
395
+ },
396
+ "APNOutputs/OEIS/oeis_278070_conjecture_0.lean": {
397
+ "group": "oeis_278070",
398
+ "rule": "OEIS sequence identifier from the filename stem"
399
+ },
400
+ "APNOutputs/OEIS/oeis_282779_conjecture_0.lean": {
401
+ "group": "oeis_282779",
402
+ "rule": "OEIS sequence identifier from the filename stem"
403
+ },
404
+ "APNOutputs/OEIS/oeis_289411_conjecture_0.lean": {
405
+ "group": "oeis_289411",
406
+ "rule": "OEIS sequence identifier from the filename stem"
407
+ },
408
+ "APNOutputs/OEIS/oeis_2897_conjecture_0.lean": {
409
+ "group": "oeis_2897",
410
+ "rule": "OEIS sequence identifier from the filename stem"
411
+ },
412
+ "APNOutputs/OEIS/oeis_306424_conjecture_0.lean": {
413
+ "group": "oeis_306424",
414
+ "rule": "OEIS sequence identifier from the filename stem"
415
+ },
416
+ "APNOutputs/OEIS/oeis_307865_conjecture_0.lean": {
417
+ "group": "oeis_307865",
418
+ "rule": "OEIS sequence identifier from the filename stem"
419
+ },
420
+ "APNOutputs/OEIS/oeis_323557_conjecture_0.lean": {
421
+ "group": "oeis_323557",
422
+ "rule": "OEIS sequence identifier from the filename stem"
423
+ },
424
+ "APNOutputs/OEIS/oeis_340737_conjecture_0.lean": {
425
+ "group": "oeis_340737",
426
+ "rule": "OEIS sequence identifier from the filename stem"
427
+ },
428
+ "APNOutputs/OEIS/oeis_341254_conjecture_0.lean": {
429
+ "group": "oeis_341254",
430
+ "rule": "OEIS sequence identifier from the filename stem"
431
+ },
432
+ "APNOutputs/OEIS/oeis_363347_conjecture_2.lean": {
433
+ "group": "oeis_363347",
434
+ "rule": "OEIS sequence identifier from the filename stem"
435
+ },
436
+ "APNOutputs/OEIS/oeis_372761_conjecture_2.lean": {
437
+ "group": "oeis_372761",
438
+ "rule": "OEIS sequence identifier from the filename stem"
439
+ },
440
+ "APNOutputs/OEIS/oeis_51293_conjecture_0.lean": {
441
+ "group": "oeis_51293",
442
+ "rule": "OEIS sequence identifier from the filename stem"
443
+ },
444
+ "APNOutputs/OEIS/oeis_62567_conjecture_0.lean": {
445
+ "group": "oeis_62567",
446
+ "rule": "OEIS sequence identifier from the filename stem"
447
+ },
448
+ "APNOutputs/OEIS/oeis_A028859_conjecture_1.lean": {
449
+ "group": "oeis_a028859",
450
+ "rule": "OEIS sequence identifier from the filename stem"
451
+ },
452
+ "APNOutputs/OEIS/oeis_A258667_conjecture_0.lean": {
453
+ "group": "oeis_a258667",
454
+ "rule": "OEIS sequence identifier from the filename stem"
455
+ },
456
+ "APNOutputs/OEIS/oeis_a211417_conjecture_specific.lean": {
457
+ "group": "oeis_a211417",
458
+ "rule": "OEIS sequence identifier from the filename stem"
459
+ },
460
+ "APNOutputs/OEIS/oeis_a237271_conjecture_2.lean": {
461
+ "group": "oeis_a237271",
462
+ "rule": "OEIS sequence identifier from the filename stem"
463
+ },
464
+ "APNOutputs/OEIS/oeis_a300997_finite_difference_is_one_or_two.lean": {
465
+ "group": "oeis_a300997",
466
+ "rule": "OEIS sequence identifier from the filename stem"
467
+ },
468
+ "APNOutputs/OEIS/oeis_a363102_conjecture_1.lean": {
469
+ "group": "oeis_a363102",
470
+ "rule": "OEIS sequence identifier from the filename stem"
471
+ },
472
+ "APNOutputs/OEIS/oeis_a368692_conjecture_integrality.lean": {
473
+ "group": "oeis_a368692",
474
+ "rule": "OEIS sequence identifier from the filename stem"
475
+ },
476
+ "APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_of_directed_colimit.lean": {
477
+ "group": "stacks_project",
478
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
479
+ },
480
+ "APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_transitive.lean": {
481
+ "group": "stacks_project",
482
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
483
+ },
484
+ "APNOutputs/StacksProject/AlgebrasOverFields_geometrically_irreducible_tfae.lean": {
485
+ "group": "stacks_project",
486
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
487
+ },
488
+ "APNOutputs/StacksProject/AlgebrasOverFields_minimalPrimes_bijection_of_geometrically_IrreducibleAlgebra.lean": {
489
+ "group": "stacks_project",
490
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
491
+ },
492
+ "APNOutputs/StacksProject/AlgebrasOverFields_unique_minimal_prime_of_tensor_product_of_separably_closed.lean": {
493
+ "group": "stacks_project",
494
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
495
+ },
496
+ "APNOutputs/StacksProject/GeometricallyReduced_isGeometricallyReduced_iff_of_separable_algebraic.lean": {
497
+ "group": "stacks_project",
498
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
499
+ },
500
+ "APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separable.lean": {
501
+ "group": "stacks_project",
502
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
503
+ },
504
+ "APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separablyGenerated.lean": {
505
+ "group": "stacks_project",
506
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
507
+ },
508
+ "APNOutputs/StacksProject/NormalRings_NormalDomain_powerSeries.lean": {
509
+ "group": "stacks_project",
510
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
511
+ },
512
+ "APNOutputs/StacksProject/NormalRings_normalRing_characterization.lean": {
513
+ "group": "stacks_project",
514
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
515
+ },
516
+ "APNOutputs/StacksProject/Separability_exists_isPurelyInseparable_isSeparablyGenerated_of_essFiniteType.lean": {
517
+ "group": "stacks_project",
518
+ "rule": "no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject"
519
+ },
520
+ "NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf": {
521
+ "group": "ai_graph_reconstruction",
522
+ "rule": "paired with APNOutputs/AICollaborator/Graphs/bipartite_graph_reconstruction_conjecture_2.lean"
523
+ },
524
+ "NaturalLanguageProofs/AICollaborator/hilbert.pdf": {
525
+ "group": "ai_hilbert_functions",
526
+ "rule": "paired with APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_*.lean"
527
+ },
528
+ "NaturalLanguageProofs/ErdosProblems/erdos125.pdf": {
529
+ "group": "erdos_125",
530
+ "rule": "erdos problem number from the filename stem"
531
+ },
532
+ "NaturalLanguageProofs/ErdosProblems/erdos12i.pdf": {
533
+ "group": "erdos_12",
534
+ "rule": "erdos problem number from the filename stem"
535
+ },
536
+ "NaturalLanguageProofs/ErdosProblems/erdos12ii.pdf": {
537
+ "group": "erdos_12",
538
+ "rule": "erdos problem number from the filename stem"
539
+ },
540
+ "NaturalLanguageProofs/ErdosProblems/erdos138.pdf": {
541
+ "group": "erdos_138",
542
+ "rule": "erdos problem number from the filename stem"
543
+ },
544
+ "NaturalLanguageProofs/ErdosProblems/erdos152.pdf": {
545
+ "group": "erdos_152",
546
+ "rule": "erdos problem number from the filename stem"
547
+ },
548
+ "NaturalLanguageProofs/ErdosProblems/erdos26.pdf": {
549
+ "group": "erdos_26",
550
+ "rule": "erdos problem number from the filename stem"
551
+ },
552
+ "NaturalLanguageProofs/ErdosProblems/erdos741i.pdf": {
553
+ "group": "erdos_741",
554
+ "rule": "erdos problem number from the filename stem"
555
+ },
556
+ "NaturalLanguageProofs/ErdosProblems/erdos741ii.pdf": {
557
+ "group": "erdos_741",
558
+ "rule": "erdos problem number from the filename stem"
559
+ },
560
+ "NaturalLanguageProofs/ErdosProblems/erdos846.pdf": {
561
+ "group": "erdos_846",
562
+ "rule": "erdos problem number from the filename stem"
563
+ },
564
+ "NaturalLanguageProofs/OEIS/oeis_51293.pdf": {
565
+ "group": "oeis_51293",
566
+ "rule": "OEIS sequence identifier from the filename stem"
567
+ }
568
+ }
569
+ }
components/google-alphaproof-nexus/audit/packing.json ADDED
@@ -0,0 +1,7 @@
 
 
 
 
 
 
 
 
1
+ {
2
+ "packer": "tools/build_google_alphaproof_nexus.py:pack_split (copied from components/openai-math)",
3
+ "tokenizer_sha256": "18b5aa749c6c791ab59c2fd1eb39dfbaa155247322119d3924d3cb9768622417",
4
+ "tokenizer_note": "SentencePiece decode is whitespace-normalizing for runs of spaces; every packed row is verified by re-encoding and by decoded-text comparison in the spot check.",
5
+ "sequence_length": 2048,
6
+ "shard_rows": 256
7
+ }
components/google-alphaproof-nexus/audit/source-files.jsonl ADDED
@@ -0,0 +1,92 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"bytes":487,"path":".github/workflows/lean_action_ci.yml","sha256":"3b0d49449c2e6d4ee04ce6e57c23dde1a836314b502351f1f6203477e2aafe8f","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/.github/workflows/lean_action_ci.yml"}
2
+ {"bytes":263,"path":".gitignore","sha256":"0f143c917804dd5bb407dc211efc324a5002488bacd878897f93dbbed6de9aa0","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/.gitignore"}
3
+ {"bytes":64249,"path":"APNOutputs/AICollaborator/AdditiveCombinatorics/57.lean","sha256":"695c74089334ab67d558d30f91c57bb8c3c7a833542204f301e0249a741fee21","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AdditiveCombinatorics/57.lean"}
4
+ {"bytes":11817,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_1.lean","sha256":"1bf1aa30c21bd21c8815ddf49d2b58e86e4acabd4d9c7035089b14ee244ea012","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_1.lean"}
5
+ {"bytes":27023,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_2.lean","sha256":"0a8959982f436121de309465cff0550c869ad3cb749d04e8dbe06ceec6d70d45","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_2.lean"}
6
+ {"bytes":75753,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_3.lean","sha256":"dbad59e1409319f51ce095392bdcf4c89919538886f5bba5bed3874fa9e963c2","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_3.lean"}
7
+ {"bytes":50626,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_4.lean","sha256":"b804d6b670c24e2412c1d2afd6aaedec1b35a0f58c614fef8c8350582061350c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_4.lean"}
8
+ {"bytes":125326,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_5.lean","sha256":"d654c3dbdb006712418b30e8c43d1de45d32ba1a5c7d514033e47066f7b2ef26","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_5.lean"}
9
+ {"bytes":36615,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_7.lean","sha256":"329a90eb36a99413bd9a78ffb81392ec4332dc5ffe7ae1672990f948ad9d8979","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_7.lean"}
10
+ {"bytes":36432,"path":"APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_8.lean","sha256":"64f77783478d32f3a5e3bdec3fcecc49ddcf012182a9efc2770abc2a5159e723","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_8.lean"}
11
+ {"bytes":39394,"path":"APNOutputs/AICollaborator/Graphs/GraphConjecture2.lean","sha256":"0ba2f8d543453911f18efd52ec3f6662d78e7a5502331708a00e1f0d9cc95d4c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/Graphs/GraphConjecture2.lean"}
12
+ {"bytes":61516,"path":"APNOutputs/AICollaborator/Graphs/bipartite_graph_reconstruction_conjecture_2.lean","sha256":"e3cf620366c07104dfcea93bb057ce4a2ea604fc1141f7e4ea5da6a6c51d555f","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/Graphs/bipartite_graph_reconstruction_conjecture_2.lean"}
13
+ {"bytes":32994,"path":"APNOutputs/AICollaborator/Optimization/LastIterateConvergence.lean","sha256":"8dbeb62e5ed2b3221b3f83e8f371491e327d8c002f03e916691e2ebb780d01c1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/Optimization/LastIterateConvergence.lean"}
14
+ {"bytes":106151,"path":"APNOutputs/AICollaborator/QuantumOptics/35.lean","sha256":"a303f0bd4b1cfc571840a102b89fe65e73f5d647c7305a22788b959d3d01eeb8","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/QuantumOptics/35.lean"}
15
+ {"bytes":31767,"path":"APNOutputs/AICollaborator/QuantumOptics/MonochromaticQuantumGraph.lean","sha256":"3d9a407f092940276d83a0518a5a25ee79a491b5352eeac5fced7c25117981dc","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/AICollaborator/QuantumOptics/MonochromaticQuantumGraph.lean"}
16
+ {"bytes":48859,"path":"APNOutputs/ErdosProblems/erdos_12.parts.i.lean","sha256":"c42238294d8d2d88ec752c5f293486a759b163412929dc4bc67d08702a499a59","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_12.parts.i.lean"}
17
+ {"bytes":36449,"path":"APNOutputs/ErdosProblems/erdos_12.parts.ii.lean","sha256":"f155135c1e845cb3041abe20edce22e660c407b48620d696ea6325c75ae91b66","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_12.parts.ii.lean"}
18
+ {"bytes":22480,"path":"APNOutputs/ErdosProblems/erdos_125.variants.positive_lower_density.lean","sha256":"12e2df1ba973bde5028ccd1f66e7ae3ba493144213337d9f71f39e28d5d0c47c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_125.variants.positive_lower_density.lean"}
19
+ {"bytes":33637,"path":"APNOutputs/ErdosProblems/erdos_138.variants.difference.lean","sha256":"c43f8170bbf340464e25fbdc479f6d407d04cda5850521e572aea2d7b53af133","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_138.variants.difference.lean"}
20
+ {"bytes":32896,"path":"APNOutputs/ErdosProblems/erdos_152.lean","sha256":"fb9cf6bb4de746f9e520e07ce9fccbddd6cd63eafb674bbe60dd457a827c6e35","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_152.lean"}
21
+ {"bytes":45108,"path":"APNOutputs/ErdosProblems/erdos_26.variants.tenenbaum.lean","sha256":"5c48abbba23217e2f5426cb43c3efbf9ab8e6e1a1fa7fde8103a7956c5939afe","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_26.variants.tenenbaum.lean"}
22
+ {"bytes":87669,"path":"APNOutputs/ErdosProblems/erdos_741.parts.i.lean","sha256":"481e87ebaa4058e5ce4f2cf85f229cf4382a633c6de02cd4617c8bf1212184df","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_741.parts.i.lean"}
23
+ {"bytes":23385,"path":"APNOutputs/ErdosProblems/erdos_741.parts.ii.lean","sha256":"77a222124ab3d8b85346656614dc5cb5f57d41c911525ca6641255b4ba9260c8","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_741.parts.ii.lean"}
24
+ {"bytes":45873,"path":"APNOutputs/ErdosProblems/erdos_846.lean","sha256":"47f77fbc9646be4c1c58ae5713c7dcbee610f8b006e049f4f6c74ebc6159d227","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/ErdosProblems/erdos_846.lean"}
25
+ {"bytes":11899,"path":"APNOutputs/OEIS/A224515_conjecture_existence.lean","sha256":"840ef4884695144094f595b872c18b6b4a6e5da24cb64691c151594fe24f5012","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/A224515_conjecture_existence.lean"}
26
+ {"bytes":29307,"path":"APNOutputs/OEIS/A309132_conjecture_carmichael.lean","sha256":"52949abaa44d13c5ccdb59c20a0726ae4af5e8f14f47cf1e26cd745775ac7c67","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/A309132_conjecture_carmichael.lean"}
27
+ {"bytes":10853,"path":"APNOutputs/OEIS/A382590_conjecture_kth_prime_factor_is_eventually_periodic.lean","sha256":"6ddbe53ac8365a9b15cece281deacc4c038190d40cc85811d3a144f964e853f6","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/A382590_conjecture_kth_prime_factor_is_eventually_periodic.lean"}
28
+ {"bytes":20338,"path":"APNOutputs/OEIS/a091669_conjecture_primitive_root.lean","sha256":"732d56d01163166ababd6036b5c3b3d881b7ab284c5cfc251bba9da34ba7541c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/a091669_conjecture_primitive_root.lean"}
29
+ {"bytes":10537,"path":"APNOutputs/OEIS/a325046_odd_terms_at_k_times_k_plus_1.lean","sha256":"41a538eba04eb86ef9817f6b1fe5b80ceeb36550a1d680bcecf9ef3477f0904e","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/a325046_odd_terms_at_k_times_k_plus_1.lean"}
30
+ {"bytes":9519,"path":"APNOutputs/OEIS/oeis_103311_conjecture_0.lean","sha256":"d419f2c4e3276ed9325a223e952b3de754f2a0f8dca3f73f40a2e859eece6cf2","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_103311_conjecture_0.lean"}
31
+ {"bytes":16499,"path":"APNOutputs/OEIS/oeis_108_conjecture_2.lean","sha256":"77cc6f79effff954d6c26c4802dc24533f65c49d848c348d75c1ccf8556430c8","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_108_conjecture_2.lean"}
32
+ {"bytes":4174,"path":"APNOutputs/OEIS/oeis_113254_conjecture_0.lean","sha256":"a3e69954964e44ad768cf10b00599bd1839ee351ac3d5041ed06ae9fd5ff949f","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_113254_conjecture_0.lean"}
33
+ {"bytes":17155,"path":"APNOutputs/OEIS/oeis_175386_conjecture_0.lean","sha256":"d3d3ac8e012f8b7646455297123cbe487963535b1fde3106adfd1b1353f43523","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_175386_conjecture_0.lean"}
34
+ {"bytes":21018,"path":"APNOutputs/OEIS/oeis_194806_conjecture_0.lean","sha256":"fbedea042aadfc835b137de110dc24fcb39e4fb911f387d2418f18ee9e652402","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_194806_conjecture_0.lean"}
35
+ {"bytes":26965,"path":"APNOutputs/OEIS/oeis_227582_conjecture_0.lean","sha256":"990c4b738924fd0cd550ec0db06f2b45c45d5fc7139d9106620c7abdc1d529cb","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_227582_conjecture_0.lean"}
36
+ {"bytes":31635,"path":"APNOutputs/OEIS/oeis_228143_conjecture_1.lean","sha256":"12b0ae8340d762e1600b4e29a823c6086fbccb319c5e99d440227eaccdb965c3","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_228143_conjecture_1.lean"}
37
+ {"bytes":7113,"path":"APNOutputs/OEIS/oeis_243106_conjecture_0.lean","sha256":"1e44a32c1eb8b735752b5d7d48b8e670e66854269c51c6ac686d722547b84dc7","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_243106_conjecture_0.lean"}
38
+ {"bytes":7791,"path":"APNOutputs/OEIS/oeis_248802_conjecture_0.lean","sha256":"ecb5a332d9db4c04a10c0b827a34c4915e2c45794102e05c7c44fcfa2ef1c94c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_248802_conjecture_0.lean"}
39
+ {"bytes":33002,"path":"APNOutputs/OEIS/oeis_248802_conjecture_4.lean","sha256":"6d7ed71c8091f9bf272a477ff2784d8f5439200f6249d3caa7ab8e34e922d957","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_248802_conjecture_4.lean"}
40
+ {"bytes":5913,"path":"APNOutputs/OEIS/oeis_256012_conjecture_0.lean","sha256":"b654ac15402dfd2a2ca4d19b7745de7386366eca074e16cbe1c8c0674974523c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_256012_conjecture_0.lean"}
41
+ {"bytes":8894,"path":"APNOutputs/OEIS/oeis_267581_conjecture_0.lean","sha256":"104cf8b11d1685cca1a309717ad77fcede56448f5fa9c2cf3f718d99ec9353db","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_267581_conjecture_0.lean"}
42
+ {"bytes":22698,"path":"APNOutputs/OEIS/oeis_271591_conjecture_0.lean","sha256":"b00b259a0508369c6d1f20c3e31ad7c3068e7aa7790e59b33a811e5359148f89","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_271591_conjecture_0.lean"}
43
+ {"bytes":3406,"path":"APNOutputs/OEIS/oeis_278070_conjecture_0.lean","sha256":"d6744aaa47d00b6144c830144747f037d258f5aed857102fb99ea0dc50e5ebb4","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_278070_conjecture_0.lean"}
44
+ {"bytes":7276,"path":"APNOutputs/OEIS/oeis_282779_conjecture_0.lean","sha256":"6eeaf9dfb51e1af35996de05398eb537835f8f333bd886fdee2b0fba674cd87e","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_282779_conjecture_0.lean"}
45
+ {"bytes":8644,"path":"APNOutputs/OEIS/oeis_289411_conjecture_0.lean","sha256":"ad288e37e9865e0fc265ddd1e35dd74529eff0c6e2e7fe9f4a38d0799ce1e370","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_289411_conjecture_0.lean"}
46
+ {"bytes":28876,"path":"APNOutputs/OEIS/oeis_2897_conjecture_0.lean","sha256":"ee1c14c2d19f0205125fe5a936574fc8dc02e198e81f9c2934fdb21e4bea8289","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_2897_conjecture_0.lean"}
47
+ {"bytes":14954,"path":"APNOutputs/OEIS/oeis_306424_conjecture_0.lean","sha256":"ee082ee70232def06a9cffb99e7e8188fdcc2db275d054e630f2981f15b9e7bd","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_306424_conjecture_0.lean"}
48
+ {"bytes":8727,"path":"APNOutputs/OEIS/oeis_307865_conjecture_0.lean","sha256":"d85aec129801dec4ba343a3c9fe525a17fe35e9ef8676a9bf62aecabb5a4a986","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_307865_conjecture_0.lean"}
49
+ {"bytes":10998,"path":"APNOutputs/OEIS/oeis_323557_conjecture_0.lean","sha256":"18b5bc4c7b634265fb3839c19b5ddd8c0d8765c27e4d583a28262b94f98a9534","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_323557_conjecture_0.lean"}
50
+ {"bytes":23411,"path":"APNOutputs/OEIS/oeis_340737_conjecture_0.lean","sha256":"7e6cf1b6965b404c965392a2049946d919455e36693cf4f04eb34655aa07b8c0","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_340737_conjecture_0.lean"}
51
+ {"bytes":6758,"path":"APNOutputs/OEIS/oeis_341254_conjecture_0.lean","sha256":"96ed7014c7c434a2af114f435bd0b708bfb3c5f368ebcd445645698b505da317","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_341254_conjecture_0.lean"}
52
+ {"bytes":38821,"path":"APNOutputs/OEIS/oeis_363347_conjecture_2.lean","sha256":"8933339b5d66f4742f9aa8a81b867eb7e60e97bb85c9d6935c2fa1131e010eb1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_363347_conjecture_2.lean"}
53
+ {"bytes":37580,"path":"APNOutputs/OEIS/oeis_372761_conjecture_2.lean","sha256":"810c0d64c6822c4b71cd1e15c69fff7f3b767b19e42755829078512f25904f29","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_372761_conjecture_2.lean"}
54
+ {"bytes":41881,"path":"APNOutputs/OEIS/oeis_51293_conjecture_0.lean","sha256":"f6e22baaff5cd8918bd57e5aa031bd5341c874bbeb7c01362ba0df1aa337cbd1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_51293_conjecture_0.lean"}
55
+ {"bytes":15635,"path":"APNOutputs/OEIS/oeis_62567_conjecture_0.lean","sha256":"63a178e1462668e6b94f663bc8532d4eae65784e16ebca8956f37e31b625452c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_62567_conjecture_0.lean"}
56
+ {"bytes":28747,"path":"APNOutputs/OEIS/oeis_A028859_conjecture_1.lean","sha256":"1495c5c563827520b926f019e0ac1c687fa27a9a90b0434bf51bb07c70de2687","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_A028859_conjecture_1.lean"}
57
+ {"bytes":28220,"path":"APNOutputs/OEIS/oeis_A258667_conjecture_0.lean","sha256":"78965da8296495fdeae1c64feb7661c29d0c307ff01462515ab7957a257ced00","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_A258667_conjecture_0.lean"}
58
+ {"bytes":12940,"path":"APNOutputs/OEIS/oeis_a211417_conjecture_specific.lean","sha256":"a9236cfb86443f76e7b3844c9a12c1139c372247d3ace331fa4f0afdcf0530d8","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_a211417_conjecture_specific.lean"}
59
+ {"bytes":4405,"path":"APNOutputs/OEIS/oeis_a237271_conjecture_2.lean","sha256":"3defe7d0314e5f6d498ac14b7821ebd0915e5a3572ce1e39e771b5b8ef9bedd3","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_a237271_conjecture_2.lean"}
60
+ {"bytes":42505,"path":"APNOutputs/OEIS/oeis_a300997_finite_difference_is_one_or_two.lean","sha256":"ce24e889e4baaffd38eab631144f09cad87c398109c2ac89e7625b8a091bf47c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_a300997_finite_difference_is_one_or_two.lean"}
61
+ {"bytes":16131,"path":"APNOutputs/OEIS/oeis_a363102_conjecture_1.lean","sha256":"ef4327bacfdc9c9680d42e9cd36e8f0a3c539bf559c47de7dee3f9eb15d57022","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_a363102_conjecture_1.lean"}
62
+ {"bytes":9060,"path":"APNOutputs/OEIS/oeis_a368692_conjecture_integrality.lean","sha256":"40d69f1d2c1f1976cc202d436542bdf6b7c431f8a8fce507037be89332e53c8e","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/OEIS/oeis_a368692_conjecture_integrality.lean"}
63
+ {"bytes":40553,"path":"APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_of_directed_colimit.lean","sha256":"d8004be3cfd5b373b7e1d0deeba66dfd1a2bdbd63ec95e5837c81b585f001d83","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_of_directed_colimit.lean"}
64
+ {"bytes":21093,"path":"APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_transitive.lean","sha256":"22ec34b14772208094efa09d963a474f1ad8a29af23ed36ec9e59f2132f4757c","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/AlgebrasOverFields_GeometricallyIrreducibleAlgebra_transitive.lean"}
65
+ {"bytes":64454,"path":"APNOutputs/StacksProject/AlgebrasOverFields_geometrically_irreducible_tfae.lean","sha256":"bab8219fe7dcda7408a264734c355d9e1fdf47fdce96eba450bbcc314b4f64a1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/AlgebrasOverFields_geometrically_irreducible_tfae.lean"}
66
+ {"bytes":27413,"path":"APNOutputs/StacksProject/AlgebrasOverFields_minimalPrimes_bijection_of_geometrically_IrreducibleAlgebra.lean","sha256":"c12fc561bf9a5587afc8a23b2bcb3c44e3d93e332b2eda07d03dc3d5b7bf15d6","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/AlgebrasOverFields_minimalPrimes_bijection_of_geometrically_IrreducibleAlgebra.lean"}
67
+ {"bytes":41095,"path":"APNOutputs/StacksProject/AlgebrasOverFields_unique_minimal_prime_of_tensor_product_of_separably_closed.lean","sha256":"7342412321e9c4fa8a14d03469ac39ae7a4e1c053ece70178f1684f832398109","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/AlgebrasOverFields_unique_minimal_prime_of_tensor_product_of_separably_closed.lean"}
68
+ {"bytes":51354,"path":"APNOutputs/StacksProject/GeometricallyReduced_isGeometricallyReduced_iff_of_separable_algebraic.lean","sha256":"d02dcbf4a97169903dbafef53a731b3f05c5e8221b428ce63dda543329a50ccf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/GeometricallyReduced_isGeometricallyReduced_iff_of_separable_algebraic.lean"}
69
+ {"bytes":22820,"path":"APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separable.lean","sha256":"09fdd23fd987306af538177ab3a14a57359a9402257bcc4176c31fed00cf16a3","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separable.lean"}
70
+ {"bytes":48185,"path":"APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separablyGenerated.lean","sha256":"5c819b5f3714d3df3ddf7ccc1c2ef95e40ca32de1cc086b39a76bcd04091a61f","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/GeometricallyReduced_isReduced_tensorProduct_of_separablyGenerated.lean"}
71
+ {"bytes":25444,"path":"APNOutputs/StacksProject/NormalRings_NormalDomain_powerSeries.lean","sha256":"04301f0af8332e06ed5bcb8b5f45763c5374f65e02b80ab3caef4ed46bcf8966","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/NormalRings_NormalDomain_powerSeries.lean"}
72
+ {"bytes":51267,"path":"APNOutputs/StacksProject/NormalRings_normalRing_characterization.lean","sha256":"4beb16ab593b0868afc238450bbdc595b1473bea53cf7a5070b4b9e043451fa3","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/NormalRings_normalRing_characterization.lean"}
73
+ {"bytes":96588,"path":"APNOutputs/StacksProject/Separability_exists_isPurelyInseparable_isSeparablyGenerated_of_essFiniteType.lean","sha256":"acda02dec42ac1bdaf7ba3338b355be6c20706b1e318e3240d4f35578deae4e1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/APNOutputs/StacksProject/Separability_exists_isPurelyInseparable_isSeparablyGenerated_of_essFiniteType.lean"}
74
+ {"bytes":970,"path":"CONTRIBUTING.md","sha256":"f6fbadc60d6a18293f9b78442f80ef2381abb35cdd95a0d071265f65b70914db","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/CONTRIBUTING.md"}
75
+ {"bytes":11358,"path":"LICENSE","sha256":"cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/LICENSE"}
76
+ {"bytes":155211,"path":"NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf","sha256":"c2b8922bc3959d0b16e0541d509f9056545b3647f573da39943bac7ce2cfaaf6","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf"}
77
+ {"bytes":166114,"path":"NaturalLanguageProofs/AICollaborator/hilbert.pdf","sha256":"fd368b619922e24153911cadfb95bc5a9cc0aa022b924e7cad22819fd3913bf6","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/AICollaborator/hilbert.pdf"}
78
+ {"bytes":94328,"path":"NaturalLanguageProofs/ErdosProblems/erdos125.pdf","sha256":"09ae6a9792a4c71cba4af5382bc478887d3dea9844e20a88086bd3bbd1f8fd09","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos125.pdf"}
79
+ {"bytes":135167,"path":"NaturalLanguageProofs/ErdosProblems/erdos12i.pdf","sha256":"2165f6786e7910bf0655a36c0fb345fcf15c4428c95abe6d2655ab6ec1f40117","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos12i.pdf"}
80
+ {"bytes":131062,"path":"NaturalLanguageProofs/ErdosProblems/erdos12ii.pdf","sha256":"79841e1a68adb697a5c0711bbbd12e4eb5eef74243528704aa847392da17cfdf","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos12ii.pdf"}
81
+ {"bytes":71009,"path":"NaturalLanguageProofs/ErdosProblems/erdos138.pdf","sha256":"bae9beb93c0c87e0e7d70a9744073a01421d85925e9b4530405d1cfc5d445df1","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos138.pdf"}
82
+ {"bytes":110988,"path":"NaturalLanguageProofs/ErdosProblems/erdos152.pdf","sha256":"ec57666d3b8c052c560848f7bd8501f34ad0fcbc00c1be7c7a9010c0c274aa86","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos152.pdf"}
83
+ {"bytes":129441,"path":"NaturalLanguageProofs/ErdosProblems/erdos26.pdf","sha256":"11f699568a86c074d32fd3e0529766ab63f6f0eefaae15c16c9a3d062f5bd929","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos26.pdf"}
84
+ {"bytes":97016,"path":"NaturalLanguageProofs/ErdosProblems/erdos741i.pdf","sha256":"62860c9e101e00969b0eed6429bc8514c7917bf7a8b8226e2bd3c7e7c7bfa164","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos741i.pdf"}
85
+ {"bytes":92877,"path":"NaturalLanguageProofs/ErdosProblems/erdos741ii.pdf","sha256":"d77702d120c9bc7703d8273275e2868c9b6d2e6e828382dac264103adad319b4","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos741ii.pdf"}
86
+ {"bytes":115400,"path":"NaturalLanguageProofs/ErdosProblems/erdos846.pdf","sha256":"5bc626735f861e310b8c62e467973c3762d48df8c3edc3737924b7799efe7a15","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/ErdosProblems/erdos846.pdf"}
87
+ {"bytes":148653,"path":"NaturalLanguageProofs/OEIS/oeis_51293.pdf","sha256":"8f464238427f434c377101fa5ac4ccb4864af62ac74bd733b4947a1c3e6247bd","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/NaturalLanguageProofs/OEIS/oeis_51293.pdf"}
88
+ {"bytes":4112,"path":"README.md","sha256":"c65d10d06ebadcb78ea0a13ca391003f6d018aa50903ac71fb871e5e946366db","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/README.md"}
89
+ {"bytes":6929,"path":"erdos_problems_attempted.txt","sha256":"144046c754a6ba494944bbd9378d26e23355fb0298ef4e78ca821ded44d4f6a3","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/erdos_problems_attempted.txt"}
90
+ {"bytes":3468,"path":"lake-manifest.json","sha256":"d5f3544a12eb2b6ae6913524e16b0e62cad4ca279d0cfa89e4d4bd5c467bffd5","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/lake-manifest.json"}
91
+ {"bytes":434,"path":"lakefile.toml","sha256":"00f7314d132e9e0ececa7f128f152253c81c25509d5fbdd08c63af3e1ad53d8a","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/lakefile.toml"}
92
+ {"bytes":25,"path":"lean-toolchain","sha256":"d55ca0039a5479db5b38919d005b2c427b89b3be4f0184a20f2f4eae931f5bdb","source_url":"https://github.com/google-deepmind/alphaproof-nexus-results/blob/0647711a71183c1ea492ad60860776617ce1ea88/lean-toolchain"}
components/google-alphaproof-nexus/audit/spotcheck.json ADDED
@@ -0,0 +1,382 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "tokenizer": {
3
+ "path": "tokenizer/tokenizer.model",
4
+ "sha256": "18b5aa749c6c791ab59c2fd1eb39dfbaa155247322119d3924d3cb9768622417",
5
+ "vocab_size": 32768,
6
+ "ids": {
7
+ "pad": 0,
8
+ "bos": 1,
9
+ "eos": 2,
10
+ "unk": 3
11
+ }
12
+ },
13
+ "checks": [
14
+ {
15
+ "name": "split_separation_from_emitted_files",
16
+ "groups_in_files": 51,
17
+ "groups_with_multiple_splits": {},
18
+ "groups_in_audit_but_absent_from_files": [],
19
+ "split_record_counts": {
20
+ "train": 80,
21
+ "dev": 1,
22
+ "test": 2
23
+ },
24
+ "split_group_counts": {
25
+ "train": 49,
26
+ "dev": 1,
27
+ "test": 1
28
+ },
29
+ "passed": true
30
+ },
31
+ {
32
+ "name": "packed_rows_decode_to_source_text",
33
+ "cases": 5,
34
+ "byte_exact_all": false,
35
+ "whitespace_normalized_all": true,
36
+ "non_whitespace_characters_identical_all": true,
37
+ "reencode_identical_all": true,
38
+ "bos_eos_all": true,
39
+ "record_types_covered": [
40
+ "formal_lean_proof",
41
+ "natural_language_proof"
42
+ ],
43
+ "passed": true
44
+ },
45
+ {
46
+ "name": "partial_rows_matching_record_token_slices",
47
+ "cases": 2,
48
+ "passed": true
49
+ },
50
+ {
51
+ "name": "parquet_to_jsonl_roundtrip",
52
+ "cases": [
53
+ {
54
+ "split": "train",
55
+ "parquet_row": 0,
56
+ "record_id": "google-alphaproof-nexus:ai_additivecombinatorics:lean:57",
57
+ "columns": [
58
+ "id",
59
+ "text",
60
+ "group",
61
+ "license",
62
+ "paper_id",
63
+ "split",
64
+ "record_type",
65
+ "source_url",
66
+ "metadata_json"
67
+ ],
68
+ "byte_identical_canonical_json": true,
69
+ "text_sha256_matches": true
70
+ },
71
+ {
72
+ "split": "train",
73
+ "parquet_row": 40,
74
+ "record_id": "google-alphaproof-nexus:oeis_256012:lean:oeis_256012_conjecture_0",
75
+ "columns": [
76
+ "id",
77
+ "text",
78
+ "group",
79
+ "license",
80
+ "paper_id",
81
+ "split",
82
+ "record_type",
83
+ "source_url",
84
+ "metadata_json"
85
+ ],
86
+ "byte_identical_canonical_json": true,
87
+ "text_sha256_matches": true
88
+ },
89
+ {
90
+ "split": "train",
91
+ "parquet_row": 79,
92
+ "record_id": "google-alphaproof-nexus:stacks_project:lean:Separability_exists_isPurelyInseparable_isSeparablyGenerated_of_essFiniteType",
93
+ "columns": [
94
+ "id",
95
+ "text",
96
+ "group",
97
+ "license",
98
+ "paper_id",
99
+ "split",
100
+ "record_type",
101
+ "source_url",
102
+ "metadata_json"
103
+ ],
104
+ "byte_identical_canonical_json": true,
105
+ "text_sha256_matches": true
106
+ },
107
+ {
108
+ "split": "dev",
109
+ "parquet_row": 0,
110
+ "record_id": "google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0",
111
+ "columns": [
112
+ "id",
113
+ "text",
114
+ "group",
115
+ "license",
116
+ "paper_id",
117
+ "split",
118
+ "record_type",
119
+ "source_url",
120
+ "metadata_json"
121
+ ],
122
+ "byte_identical_canonical_json": true,
123
+ "text_sha256_matches": true
124
+ },
125
+ {
126
+ "split": "dev",
127
+ "parquet_row": 0,
128
+ "record_id": "google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0",
129
+ "columns": [
130
+ "id",
131
+ "text",
132
+ "group",
133
+ "license",
134
+ "paper_id",
135
+ "split",
136
+ "record_type",
137
+ "source_url",
138
+ "metadata_json"
139
+ ],
140
+ "byte_identical_canonical_json": true,
141
+ "text_sha256_matches": true
142
+ },
143
+ {
144
+ "split": "dev",
145
+ "parquet_row": 0,
146
+ "record_id": "google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0",
147
+ "columns": [
148
+ "id",
149
+ "text",
150
+ "group",
151
+ "license",
152
+ "paper_id",
153
+ "split",
154
+ "record_type",
155
+ "source_url",
156
+ "metadata_json"
157
+ ],
158
+ "byte_identical_canonical_json": true,
159
+ "text_sha256_matches": true
160
+ },
161
+ {
162
+ "split": "test",
163
+ "parquet_row": 0,
164
+ "record_id": "google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2",
165
+ "columns": [
166
+ "id",
167
+ "text",
168
+ "group",
169
+ "license",
170
+ "paper_id",
171
+ "split",
172
+ "record_type",
173
+ "source_url",
174
+ "metadata_json"
175
+ ],
176
+ "byte_identical_canonical_json": true,
177
+ "text_sha256_matches": true
178
+ },
179
+ {
180
+ "split": "test",
181
+ "parquet_row": 1,
182
+ "record_id": "google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction",
183
+ "columns": [
184
+ "id",
185
+ "text",
186
+ "group",
187
+ "license",
188
+ "paper_id",
189
+ "split",
190
+ "record_type",
191
+ "source_url",
192
+ "metadata_json"
193
+ ],
194
+ "byte_identical_canonical_json": true,
195
+ "text_sha256_matches": true
196
+ },
197
+ {
198
+ "split": "test",
199
+ "parquet_row": 1,
200
+ "record_id": "google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction",
201
+ "columns": [
202
+ "id",
203
+ "text",
204
+ "group",
205
+ "license",
206
+ "paper_id",
207
+ "split",
208
+ "record_type",
209
+ "source_url",
210
+ "metadata_json"
211
+ ],
212
+ "byte_identical_canonical_json": true,
213
+ "text_sha256_matches": true
214
+ }
215
+ ],
216
+ "passed": true
217
+ },
218
+ {
219
+ "name": "cross_split_text_uniqueness",
220
+ "records": 83,
221
+ "unique_text_hashes": 83,
222
+ "passed": true
223
+ }
224
+ ],
225
+ "decode_cases": [
226
+ {
227
+ "split": "train",
228
+ "packed_row": 219,
229
+ "column_start": 447,
230
+ "column_end": 1396,
231
+ "record_id": "google-alphaproof-nexus:erdos_741:nl:erdos741i",
232
+ "record_type": "natural_language_proof",
233
+ "tokens": 949,
234
+ "bos_ok": true,
235
+ "eos_ok": true,
236
+ "decoded_reencodes_to_same_ids": true,
237
+ "byte_exact_text_match": false,
238
+ "whitespace_normalized_text_match": true,
239
+ "decoded_characters": 2689,
240
+ "source_characters": 2695,
241
+ "whitespace_removed_text_match": true,
242
+ "decoded_preview": "Theorem. (Erd˝os #741(i)) Let A ⊆ N be such that A + A has positive upper\\ndensity. Can one always de"
243
+ },
244
+ {
245
+ "split": "train",
246
+ "packed_row": 231,
247
+ "column_start": 383,
248
+ "column_end": 1539,
249
+ "record_id": "google-alphaproof-nexus:erdos_846:nl:erdos846",
250
+ "record_type": "natural_language_proof",
251
+ "tokens": 1156,
252
+ "bos_ok": true,
253
+ "eos_ok": true,
254
+ "decoded_reencodes_to_same_ids": true,
255
+ "byte_exact_text_match": true,
256
+ "whitespace_normalized_text_match": true,
257
+ "decoded_characters": 3950,
258
+ "source_characters": 3950,
259
+ "whitespace_removed_text_match": true,
260
+ "decoded_preview": "Theorem. (Erd˝os #846) Let A ⊂ R2 be an infinite set for which there exists\\nsome ϵ > 0 such that in "
261
+ },
262
+ {
263
+ "split": "train",
264
+ "packed_row": 160,
265
+ "column_start": 284,
266
+ "column_end": 1883,
267
+ "record_id": "google-alphaproof-nexus:erdos_12:nl:erdos12i",
268
+ "record_type": "natural_language_proof",
269
+ "tokens": 1599,
270
+ "bos_ok": true,
271
+ "eos_ok": true,
272
+ "decoded_reencodes_to_same_ids": true,
273
+ "byte_exact_text_match": false,
274
+ "whitespace_normalized_text_match": true,
275
+ "decoded_characters": 4248,
276
+ "source_characters": 4253,
277
+ "whitespace_removed_text_match": true,
278
+ "decoded_preview": "Theorem. (Erd˝os #12(i)) There is an infinite set A ⊆ N satisfying that there\\nare no distinct a, b, "
279
+ },
280
+ {
281
+ "split": "train",
282
+ "packed_row": 331,
283
+ "column_start": 196,
284
+ "column_end": 1993,
285
+ "record_id": "google-alphaproof-nexus:oeis_51293:nl:oeis_51293",
286
+ "record_type": "natural_language_proof",
287
+ "tokens": 1797,
288
+ "bos_ok": true,
289
+ "eos_ok": true,
290
+ "decoded_reencodes_to_same_ids": true,
291
+ "byte_exact_text_match": false,
292
+ "whitespace_normalized_text_match": true,
293
+ "decoded_characters": 4464,
294
+ "source_characters": 4519,
295
+ "whitespace_removed_text_match": true,
296
+ "decoded_preview": "Theorem. (A conjecture of OEIS A051293, 2002) Let an denote the\\nnumber of nonempty subsets of {1, 2,"
297
+ },
298
+ {
299
+ "split": "train",
300
+ "packed_row": 277,
301
+ "column_start": 391,
302
+ "column_end": 1760,
303
+ "record_id": "google-alphaproof-nexus:oeis_278070:lean:oeis_278070_conjecture_0",
304
+ "record_type": "formal_lean_proof",
305
+ "tokens": 1369,
306
+ "bos_ok": true,
307
+ "eos_ok": true,
308
+ "decoded_reencodes_to_same_ids": true,
309
+ "byte_exact_text_match": false,
310
+ "whitespace_normalized_text_match": true,
311
+ "decoded_characters": 3350,
312
+ "source_characters": 3373,
313
+ "whitespace_removed_text_match": true,
314
+ "decoded_preview": "/-\\nCopyright 2026 Google LLC\\n\\nLicensed under the Apache License, Version 2.0 (the \"License\");\\nyou ma"
315
+ }
316
+ ],
317
+ "partial_row_cases": [
318
+ {
319
+ "split": "dev",
320
+ "packed_row": 1,
321
+ "record_id": "google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0",
322
+ "record_type": "formal_lean_proof",
323
+ "records_in_row": 1,
324
+ "tokens": 1096,
325
+ "record_token_slice": [
326
+ 2048,
327
+ 3144
328
+ ],
329
+ "packed_tokens_equal_record_token_slice": true,
330
+ "decoded_text_equal_expected_slice_decode": true,
331
+ "decoded_characters": 2119,
332
+ "source_characters": 6805,
333
+ "whitespace_removed_text_match": true,
334
+ "decoded_preview": "d\\n contradiction\\n | succ N' =>\\n have hN : N' + 1 > 0 := Nat.zero_lt_succ N'\\n have step := isGood_ste"
335
+ },
336
+ {
337
+ "split": "test",
338
+ "packed_row": 13,
339
+ "record_id": "google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction",
340
+ "record_type": "natural_language_proof",
341
+ "records_in_row": 1,
342
+ "tokens": 1787,
343
+ "record_token_slice": [
344
+ 1127,
345
+ 2914
346
+ ],
347
+ "packed_tokens_equal_record_token_slice": true,
348
+ "decoded_text_equal_expected_slice_decode": true,
349
+ "decoded_characters": 4757,
350
+ "source_characters": 7842,
351
+ "whitespace_removed_text_match": true,
352
+ "decoded_preview": "ex had degree 0,\\nthen G would not be connected; if a vertex x had the unique neighbor y, then\\nthe in"
353
+ }
354
+ ],
355
+ "status": "passed",
356
+ "measured": {
357
+ "records": 83,
358
+ "groups": 51,
359
+ "splits": {
360
+ "train": 80,
361
+ "dev": 1,
362
+ "test": 2
363
+ },
364
+ "packed": {
365
+ "train": {
366
+ "rows": 491,
367
+ "tokens": 1004469,
368
+ "loss_bearing_labels": 1004389
369
+ },
370
+ "dev": {
371
+ "rows": 2,
372
+ "tokens": 3144,
373
+ "loss_bearing_labels": 3143
374
+ },
375
+ "test": {
376
+ "rows": 14,
377
+ "tokens": 28411,
378
+ "loss_bearing_labels": 28409
379
+ }
380
+ }
381
+ }
382
+ }
components/google-alphaproof-nexus/data/dev.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6e4add453f141390f86817f7be7e547d3a5805afe163968856fd62c6e5717d20
3
+ size 3274
components/google-alphaproof-nexus/data/dev.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b403c8d1064e1b37cc120c201eb274f52c650519b5fa99aa0525109a9770d42
3
+ size 9529
components/google-alphaproof-nexus/data/test.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:438add97c80945c8e14c23056b135abde41d43cad5ab3f75dbf77dd1a038ccea
3
+ size 17741
components/google-alphaproof-nexus/data/test.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0e5668d50700c36f0875ec7e67186b5d076c57977bfd309bf1586ec2ace214c
3
+ size 27175
components/google-alphaproof-nexus/data/train.jsonl.gz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52b789c38529ec57788d4758bfc33a24d60bb93faa149b9da69a19081646e84b
3
+ size 506806
components/google-alphaproof-nexus/data/train.parquet ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:de4d65db7391f74a3a45b0b55c011c1c6469398e698799fcaabf64bc3a3bb2fd
3
+ size 554482
components/google-alphaproof-nexus/manifest.json ADDED
@@ -0,0 +1,205 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "cayley-contribution-file-manifest-v1",
3
+ "excludes": [
4
+ ".DS_Store",
5
+ "manifest.json",
6
+ "validation.json"
7
+ ],
8
+ "files": [
9
+ {
10
+ "path": "LICENSE",
11
+ "bytes": 11358,
12
+ "sha256": "cfc7749b96f63bd31c3c42b5c471bf756814053e847c10f3eb003417bc523d30"
13
+ },
14
+ {
15
+ "path": "README.md",
16
+ "bytes": 7455,
17
+ "sha256": "b91f3ec3fbb119e84aded9530371fce5be78387bab0f47856811eb2c1409c785"
18
+ },
19
+ {
20
+ "path": "audit/excluded-duplicates.jsonl",
21
+ "bytes": 0,
22
+ "sha256": "e3b0c44298fc1c149afbf4c8996fb92427ae41e4649b934ca495991b7852b855"
23
+ },
24
+ {
25
+ "path": "audit/extraction.jsonl",
26
+ "bytes": 5956,
27
+ "sha256": "7f54de79e4fb8e4fc68e26ce6f3ce964504fae8397949c9fcf5259fc3c6c8b5c"
28
+ },
29
+ {
30
+ "path": "audit/group-splits.json",
31
+ "bytes": 21947,
32
+ "sha256": "4d1ba58bd04778848b511d1df1afce534a5a88ae243f9b7331fccc0a4e76e0a1"
33
+ },
34
+ {
35
+ "path": "audit/packing.json",
36
+ "bytes": 427,
37
+ "sha256": "ca7a208fc453bbbf6381ce1cca5448c6fd2db33dc87fd792ce57da50568aa5fa"
38
+ },
39
+ {
40
+ "path": "audit/source-files.jsonl",
41
+ "bytes": 30307,
42
+ "sha256": "3d1af2f339ebc7a967365d448e1354789726bb671e285fefd1d207ecf9d4a313"
43
+ },
44
+ {
45
+ "path": "audit/spotcheck.json",
46
+ "bytes": 11262,
47
+ "sha256": "dadf695620f99e4c1085df723bc33980758cd8d36bb25d4d621115fd9ae0603e"
48
+ },
49
+ {
50
+ "path": "data/dev.jsonl.gz",
51
+ "bytes": 3274,
52
+ "sha256": "6e4add453f141390f86817f7be7e547d3a5805afe163968856fd62c6e5717d20"
53
+ },
54
+ {
55
+ "path": "data/dev.parquet",
56
+ "bytes": 9529,
57
+ "sha256": "9b403c8d1064e1b37cc120c201eb274f52c650519b5fa99aa0525109a9770d42"
58
+ },
59
+ {
60
+ "path": "data/test.jsonl.gz",
61
+ "bytes": 17741,
62
+ "sha256": "438add97c80945c8e14c23056b135abde41d43cad5ab3f75dbf77dd1a038ccea"
63
+ },
64
+ {
65
+ "path": "data/test.parquet",
66
+ "bytes": 27175,
67
+ "sha256": "c0e5668d50700c36f0875ec7e67186b5d076c57977bfd309bf1586ec2ace214c"
68
+ },
69
+ {
70
+ "path": "data/train.jsonl.gz",
71
+ "bytes": 506806,
72
+ "sha256": "52b789c38529ec57788d4758bfc33a24d60bb93faa149b9da69a19081646e84b"
73
+ },
74
+ {
75
+ "path": "data/train.parquet",
76
+ "bytes": 554482,
77
+ "sha256": "de4d65db7391f74a3a45b0b55c011c1c6469398e698799fcaabf64bc3a3bb2fd"
78
+ },
79
+ {
80
+ "path": "packed/dev-000000-input_ids.npy",
81
+ "bytes": 16512,
82
+ "sha256": "cd1b45c410121ada1e741949e32fca121b6e112bd650187fef0a3345607f68ad"
83
+ },
84
+ {
85
+ "path": "packed/dev-000000-labels.npy",
86
+ "bytes": 16512,
87
+ "sha256": "5766574e4e4c7f59d4fbb9a74a7ca917cbb3d1af3dd1e816c64dea53a2bea70c"
88
+ },
89
+ {
90
+ "path": "packed/dev-000000-role_ids.npy",
91
+ "bytes": 16512,
92
+ "sha256": "835fdd2d7f541597713b82d139b6cd7c8ff36213ac0feb8b4ead436621afd4d5"
93
+ },
94
+ {
95
+ "path": "packed/dev-records.jsonl",
96
+ "bytes": 196,
97
+ "sha256": "9bbea0240630261c5ad0c17a03e642594736e7d698e21eae9a2940a21b5f161b"
98
+ },
99
+ {
100
+ "path": "packed/dev-rows.jsonl",
101
+ "bytes": 504,
102
+ "sha256": "e897fb977bc0c365b7409f56d081dabdb4d7432ab0710e4b103bdd1e7c392b36"
103
+ },
104
+ {
105
+ "path": "packed/manifest.json",
106
+ "bytes": 5098,
107
+ "sha256": "ce1f0e5e5e6c8a440aa8b68dea3ab04f6d2c8d70862f7caa0bd558f9ed4a0b41"
108
+ },
109
+ {
110
+ "path": "packed/test-000000-input_ids.npy",
111
+ "bytes": 114816,
112
+ "sha256": "5acd5f7a62ec7269a76483f872707346ea6382b873418fef46aec168e5fcdf1c"
113
+ },
114
+ {
115
+ "path": "packed/test-000000-labels.npy",
116
+ "bytes": 114816,
117
+ "sha256": "c4c08d6f471c05f2fc95362109a5951a7a1dae6b878109360bd6e2b93ee32865"
118
+ },
119
+ {
120
+ "path": "packed/test-000000-role_ids.npy",
121
+ "bytes": 114816,
122
+ "sha256": "8a6f775404960d69d59f3748a92fb4a921ca566cd83da5b33dc26e6588cb1958"
123
+ },
124
+ {
125
+ "path": "packed/test-records.jsonl",
126
+ "bytes": 454,
127
+ "sha256": "359cbf559321e1d39f1af4d752bb0a0838dc94cb64a0d298146e29e33a49a729"
128
+ },
129
+ {
130
+ "path": "packed/test-rows.jsonl",
131
+ "bytes": 4365,
132
+ "sha256": "c1baffe8bfa404ec188e5bba4c6f31727c60bdba901599605578af04077b5040"
133
+ },
134
+ {
135
+ "path": "packed/train-000000-input_ids.npy",
136
+ "bytes": 2097280,
137
+ "sha256": "003c13d9fdebd5a069b67e04aa0cd03490e0fa0e010722dbe31c9fda6e6961b4"
138
+ },
139
+ {
140
+ "path": "packed/train-000000-labels.npy",
141
+ "bytes": 2097280,
142
+ "sha256": "d5b04f2f1d076f9b02e3def1a562514f3506ead8339ed9226cfeb43c78c15490"
143
+ },
144
+ {
145
+ "path": "packed/train-000000-role_ids.npy",
146
+ "bytes": 2097280,
147
+ "sha256": "e8af22599bd2b950a4c4caafb09c863f5ab033a9ef09e030d08cce3c9e55506b"
148
+ },
149
+ {
150
+ "path": "packed/train-000001-input_ids.npy",
151
+ "bytes": 1925248,
152
+ "sha256": "92a2886a0f4d1a022a7587345cfac4e4603ea136a7ab2ff91e22b0b945ebaa10"
153
+ },
154
+ {
155
+ "path": "packed/train-000001-labels.npy",
156
+ "bytes": 1925248,
157
+ "sha256": "0f5c56edb4aa72c7b3fc133ffb1de5d50d984e48f9311637fc2ae4a469cb6866"
158
+ },
159
+ {
160
+ "path": "packed/train-000001-role_ids.npy",
161
+ "bytes": 1925248,
162
+ "sha256": "8662942d8fb64121b854d112befc0a8037a91c7ef0140bac5e33acc507ffd352"
163
+ },
164
+ {
165
+ "path": "packed/train-records.jsonl",
166
+ "bytes": 16105,
167
+ "sha256": "9941bba85d08a75d06d3ec6575da625204b0b040919e45c87b572445d8175857"
168
+ },
169
+ {
170
+ "path": "packed/train-rows.jsonl",
171
+ "bytes": 147244,
172
+ "sha256": "7a3543f3b08a35cff7ee6a381492dedc490147c917bce51fb684bb1e9ebd0240"
173
+ },
174
+ {
175
+ "path": "sources.json",
176
+ "bytes": 1482,
177
+ "sha256": "e87a8d4f7369a22531b03b5d8b64e59c11af90690788db3f074d12f3e2c677c9"
178
+ },
179
+ {
180
+ "path": "statistics.json",
181
+ "bytes": 1324,
182
+ "sha256": "dfe435b02e026996e34c93e1ed47b238131aa03b1b35418c620a056ee4fa345d"
183
+ },
184
+ {
185
+ "path": "tokenizer/tokenizer.model",
186
+ "bytes": 503703,
187
+ "sha256": "18b5aa749c6c791ab59c2fd1eb39dfbaa155247322119d3924d3cb9768622417"
188
+ },
189
+ {
190
+ "path": "tools/build_google_alphaproof_nexus.py",
191
+ "bytes": 30947,
192
+ "sha256": "5072ffc6f9c56d554c21cf8cc1e874424df0523c547a9b3ce8f76b04cac1d359"
193
+ },
194
+ {
195
+ "path": "tools/spotcheck_google_alphaproof_nexus.py",
196
+ "bytes": 12180,
197
+ "sha256": "92983f114b4934abdd1cd0462f71e0d6558c3030b4d02f16dde26bbff7dffc4a"
198
+ },
199
+ {
200
+ "path": "tools/validate_google_alphaproof_nexus.py",
201
+ "bytes": 17043,
202
+ "sha256": "d91a7014e21b8539f56694f6fe56a3abec70040305bf850d413c40232ae6ad06"
203
+ }
204
+ ]
205
+ }
components/google-alphaproof-nexus/packed/dev-000000-input_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cd1b45c410121ada1e741949e32fca121b6e112bd650187fef0a3345607f68ad
3
+ size 16512
components/google-alphaproof-nexus/packed/dev-000000-labels.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5766574e4e4c7f59d4fbb9a74a7ca917cbb3d1af3dd1e816c64dea53a2bea70c
3
+ size 16512
components/google-alphaproof-nexus/packed/dev-000000-role_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:835fdd2d7f541597713b82d139b6cd7c8ff36213ac0feb8b4ead436621afd4d5
3
+ size 16512
components/google-alphaproof-nexus/packed/dev-records.jsonl ADDED
@@ -0,0 +1 @@
 
 
1
+ {"id":"google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0","paper_id":"oeis_243106","token_sha256":"7af575ad66d4fa68e896e924b40d9bda2713bf85a068f1ac3ba573405102b90b","tokens":3144}
components/google-alphaproof-nexus/packed/dev-rows.jsonl ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ {"length":2048,"row":0,"segments":[{"paper_id":"oeis_243106","record_end":2048,"record_id":"google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0","record_start":0,"row_end":2048,"row_start":0}],"shard":0,"shard_row":0,"stream_start":0}
2
+ {"length":1096,"row":1,"segments":[{"paper_id":"oeis_243106","record_end":3144,"record_id":"google-alphaproof-nexus:oeis_243106:lean:oeis_243106_conjecture_0","record_start":2048,"row_end":1096,"row_start":0}],"shard":0,"shard_row":1,"stream_start":2048}
components/google-alphaproof-nexus/packed/manifest.json ADDED
@@ -0,0 +1,196 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "format": "mmla-type2-packed-v1",
3
+ "phase": "pretrain",
4
+ "layout": "lm",
5
+ "seq_len": 2048,
6
+ "tokenizer": {
7
+ "kind": "sentencepiece",
8
+ "vocab_size": 32768,
9
+ "sha256": "18b5aa749c6c791ab59c2fd1eb39dfbaa155247322119d3924d3cb9768622417",
10
+ "path": "../tokenizer/tokenizer.model",
11
+ "ids": {
12
+ "pad": 0,
13
+ "bos": 1,
14
+ "eos": 2,
15
+ "unk": 3,
16
+ "boundary": 4
17
+ }
18
+ },
19
+ "quota_units": "tokens",
20
+ "source": "google-alphaproof-nexus",
21
+ "splits": {
22
+ "train": [
23
+ {
24
+ "count": 256,
25
+ "fields": {
26
+ "input_ids": {
27
+ "path": "train-000000-input_ids.npy",
28
+ "sha256": "003c13d9fdebd5a069b67e04aa0cd03490e0fa0e010722dbe31c9fda6e6961b4",
29
+ "bytes": 2097280,
30
+ "dtype": "int32",
31
+ "shape": [
32
+ 256,
33
+ 2048
34
+ ]
35
+ },
36
+ "labels": {
37
+ "path": "train-000000-labels.npy",
38
+ "sha256": "d5b04f2f1d076f9b02e3def1a562514f3506ead8339ed9226cfeb43c78c15490",
39
+ "bytes": 2097280,
40
+ "dtype": "int32",
41
+ "shape": [
42
+ 256,
43
+ 2048
44
+ ]
45
+ },
46
+ "role_ids": {
47
+ "path": "train-000000-role_ids.npy",
48
+ "sha256": "e8af22599bd2b950a4c4caafb09c863f5ab033a9ef09e030d08cce3c9e55506b",
49
+ "bytes": 2097280,
50
+ "dtype": "int32",
51
+ "shape": [
52
+ 256,
53
+ 2048
54
+ ]
55
+ }
56
+ }
57
+ },
58
+ {
59
+ "count": 235,
60
+ "fields": {
61
+ "input_ids": {
62
+ "path": "train-000001-input_ids.npy",
63
+ "sha256": "92a2886a0f4d1a022a7587345cfac4e4603ea136a7ab2ff91e22b0b945ebaa10",
64
+ "bytes": 1925248,
65
+ "dtype": "int32",
66
+ "shape": [
67
+ 235,
68
+ 2048
69
+ ]
70
+ },
71
+ "labels": {
72
+ "path": "train-000001-labels.npy",
73
+ "sha256": "0f5c56edb4aa72c7b3fc133ffb1de5d50d984e48f9311637fc2ae4a469cb6866",
74
+ "bytes": 1925248,
75
+ "dtype": "int32",
76
+ "shape": [
77
+ 235,
78
+ 2048
79
+ ]
80
+ },
81
+ "role_ids": {
82
+ "path": "train-000001-role_ids.npy",
83
+ "sha256": "8662942d8fb64121b854d112befc0a8037a91c7ef0140bac5e33acc507ffd352",
84
+ "bytes": 1925248,
85
+ "dtype": "int32",
86
+ "shape": [
87
+ 235,
88
+ 2048
89
+ ]
90
+ }
91
+ }
92
+ }
93
+ ],
94
+ "dev": [
95
+ {
96
+ "count": 2,
97
+ "fields": {
98
+ "input_ids": {
99
+ "path": "dev-000000-input_ids.npy",
100
+ "sha256": "cd1b45c410121ada1e741949e32fca121b6e112bd650187fef0a3345607f68ad",
101
+ "bytes": 16512,
102
+ "dtype": "int32",
103
+ "shape": [
104
+ 2,
105
+ 2048
106
+ ]
107
+ },
108
+ "labels": {
109
+ "path": "dev-000000-labels.npy",
110
+ "sha256": "5766574e4e4c7f59d4fbb9a74a7ca917cbb3d1af3dd1e816c64dea53a2bea70c",
111
+ "bytes": 16512,
112
+ "dtype": "int32",
113
+ "shape": [
114
+ 2,
115
+ 2048
116
+ ]
117
+ },
118
+ "role_ids": {
119
+ "path": "dev-000000-role_ids.npy",
120
+ "sha256": "835fdd2d7f541597713b82d139b6cd7c8ff36213ac0feb8b4ead436621afd4d5",
121
+ "bytes": 16512,
122
+ "dtype": "int32",
123
+ "shape": [
124
+ 2,
125
+ 2048
126
+ ]
127
+ }
128
+ }
129
+ }
130
+ ],
131
+ "test": [
132
+ {
133
+ "count": 14,
134
+ "fields": {
135
+ "input_ids": {
136
+ "path": "test-000000-input_ids.npy",
137
+ "sha256": "5acd5f7a62ec7269a76483f872707346ea6382b873418fef46aec168e5fcdf1c",
138
+ "bytes": 114816,
139
+ "dtype": "int32",
140
+ "shape": [
141
+ 14,
142
+ 2048
143
+ ]
144
+ },
145
+ "labels": {
146
+ "path": "test-000000-labels.npy",
147
+ "sha256": "c4c08d6f471c05f2fc95362109a5951a7a1dae6b878109360bd6e2b93ee32865",
148
+ "bytes": 114816,
149
+ "dtype": "int32",
150
+ "shape": [
151
+ 14,
152
+ 2048
153
+ ]
154
+ },
155
+ "role_ids": {
156
+ "path": "test-000000-role_ids.npy",
157
+ "sha256": "8a6f775404960d69d59f3748a92fb4a921ca566cd83da5b33dc26e6588cb1958",
158
+ "bytes": 114816,
159
+ "dtype": "int32",
160
+ "shape": [
161
+ 14,
162
+ 2048
163
+ ]
164
+ }
165
+ }
166
+ }
167
+ ]
168
+ },
169
+ "statistics": {
170
+ "train": {
171
+ "rows": 491,
172
+ "tokens": 1004469,
173
+ "loss_bearing_labels": 1004389
174
+ },
175
+ "dev": {
176
+ "rows": 2,
177
+ "tokens": 3144,
178
+ "loss_bearing_labels": 3143
179
+ },
180
+ "test": {
181
+ "rows": 14,
182
+ "tokens": 28411,
183
+ "loss_bearing_labels": 28409
184
+ }
185
+ },
186
+ "packing": {
187
+ "stride": 2048,
188
+ "bos_per_record": true,
189
+ "eos_per_record": true,
190
+ "mask_target_bos": true,
191
+ "mask_final_target": true,
192
+ "loss_mask": -100,
193
+ "row_attribution": "{split}-rows.jsonl",
194
+ "record_tokens": "{split}-records.jsonl"
195
+ }
196
+ }
components/google-alphaproof-nexus/packed/test-000000-input_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5acd5f7a62ec7269a76483f872707346ea6382b873418fef46aec168e5fcdf1c
3
+ size 114816
components/google-alphaproof-nexus/packed/test-000000-labels.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c4c08d6f471c05f2fc95362109a5951a7a1dae6b878109360bd6e2b93ee32865
3
+ size 114816
components/google-alphaproof-nexus/packed/test-000000-role_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a6f775404960d69d59f3748a92fb4a921ca566cd83da5b33dc26e6588cb1958
3
+ size 114816
components/google-alphaproof-nexus/packed/test-records.jsonl ADDED
@@ -0,0 +1,2 @@
 
 
 
1
+ {"id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","paper_id":"ai_graph_reconstruction","token_sha256":"dbfcc1c2d28fea7b7894fc9e454895ca224298564dc444f2d3eae0a1bf43005e","tokens":25497}
2
+ {"id":"google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction","paper_id":"ai_graph_reconstruction","token_sha256":"634acd156402c62b15f64dd658a08ffca26dbd97a7e5bdc6a4f3a56003b6793d","tokens":2914}
components/google-alphaproof-nexus/packed/test-rows.jsonl ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"length":2048,"row":0,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":2048,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":0,"row_end":2048,"row_start":0}],"shard":0,"shard_row":0,"stream_start":0}
2
+ {"length":2048,"row":1,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":4096,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":2048,"row_end":2048,"row_start":0}],"shard":0,"shard_row":1,"stream_start":2048}
3
+ {"length":2048,"row":2,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":6144,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":4096,"row_end":2048,"row_start":0}],"shard":0,"shard_row":2,"stream_start":4096}
4
+ {"length":2048,"row":3,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":8192,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":6144,"row_end":2048,"row_start":0}],"shard":0,"shard_row":3,"stream_start":6144}
5
+ {"length":2048,"row":4,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":10240,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":8192,"row_end":2048,"row_start":0}],"shard":0,"shard_row":4,"stream_start":8192}
6
+ {"length":2048,"row":5,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":12288,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":10240,"row_end":2048,"row_start":0}],"shard":0,"shard_row":5,"stream_start":10240}
7
+ {"length":2048,"row":6,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":14336,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":12288,"row_end":2048,"row_start":0}],"shard":0,"shard_row":6,"stream_start":12288}
8
+ {"length":2048,"row":7,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":16384,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":14336,"row_end":2048,"row_start":0}],"shard":0,"shard_row":7,"stream_start":14336}
9
+ {"length":2048,"row":8,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":18432,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":16384,"row_end":2048,"row_start":0}],"shard":0,"shard_row":8,"stream_start":16384}
10
+ {"length":2048,"row":9,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":20480,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":18432,"row_end":2048,"row_start":0}],"shard":0,"shard_row":9,"stream_start":18432}
11
+ {"length":2048,"row":10,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":22528,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":20480,"row_end":2048,"row_start":0}],"shard":0,"shard_row":10,"stream_start":20480}
12
+ {"length":2048,"row":11,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":24576,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":22528,"row_end":2048,"row_start":0}],"shard":0,"shard_row":11,"stream_start":22528}
13
+ {"length":2048,"row":12,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":25497,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:lean:bipartite_graph_reconstruction_conjecture_2","record_start":24576,"row_end":921,"row_start":0},{"paper_id":"ai_graph_reconstruction","record_end":1127,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction","record_start":0,"row_end":2048,"row_start":921}],"shard":0,"shard_row":12,"stream_start":24576}
14
+ {"length":1787,"row":13,"segments":[{"paper_id":"ai_graph_reconstruction","record_end":2914,"record_id":"google-alphaproof-nexus:ai_graph_reconstruction:nl:graph_reconstruction","record_start":1127,"row_end":1787,"row_start":0}],"shard":0,"shard_row":13,"stream_start":26624}
components/google-alphaproof-nexus/packed/train-000000-input_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:003c13d9fdebd5a069b67e04aa0cd03490e0fa0e010722dbe31c9fda6e6961b4
3
+ size 2097280
components/google-alphaproof-nexus/packed/train-000000-labels.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5b04f2f1d076f9b02e3def1a562514f3506ead8339ed9226cfeb43c78c15490
3
+ size 2097280
components/google-alphaproof-nexus/packed/train-000000-role_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e8af22599bd2b950a4c4caafb09c863f5ab033a9ef09e030d08cce3c9e55506b
3
+ size 2097280
components/google-alphaproof-nexus/packed/train-000001-input_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92a2886a0f4d1a022a7587345cfac4e4603ea136a7ab2ff91e22b0b945ebaa10
3
+ size 1925248
components/google-alphaproof-nexus/packed/train-000001-labels.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f5c56edb4aa72c7b3fc133ffb1de5d50d984e48f9311637fc2ae4a469cb6866
3
+ size 1925248
components/google-alphaproof-nexus/packed/train-000001-role_ids.npy ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8662942d8fb64121b854d112befc0a8037a91c7ef0140bac5e33acc507ffd352
3
+ size 1925248
components/google-alphaproof-nexus/packed/train-records.jsonl ADDED
@@ -0,0 +1,80 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {"id":"google-alphaproof-nexus:ai_additivecombinatorics:lean:57","paper_id":"ai_additivecombinatorics","token_sha256":"9aa126d213aa902192043e3140fff8035371ff78203b5a7a01671525d8c29a73","tokens":30948}
2
+ {"id":"google-alphaproof-nexus:ai_graphs:lean:GraphConjecture2","paper_id":"ai_graphs","token_sha256":"931ce04e0278d4977fc3955948e6f8df19e73d2614003f5e6214c1a281b851a2","tokens":16989}
3
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_1","paper_id":"ai_hilbert_functions","token_sha256":"1b185a73bb9d78e12a1337817140c07d06ba0e92ec5c813942fa7746a9cefefe","tokens":4801}
4
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_2","paper_id":"ai_hilbert_functions","token_sha256":"d29dc613725f6d6fc71a25e60941474502f14c3070f73496510d7ed6bc7a4001","tokens":12136}
5
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_3","paper_id":"ai_hilbert_functions","token_sha256":"ed3bf9872b04d3bee436b27bf30f7893246f0ca48c7045d92bb878a8d25ab7eb","tokens":31883}
6
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_4","paper_id":"ai_hilbert_functions","token_sha256":"6eb385224d758fd0be782b926aa5b3993b8463fe319bfca4f851839e4f9a94a8","tokens":22173}
7
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_5","paper_id":"ai_hilbert_functions","token_sha256":"16fff08270da104e7f4d1b350ca9a15573e839f142db0313d27a452769a716a9","tokens":60774}
8
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_7","paper_id":"ai_hilbert_functions","token_sha256":"75a09887d931b644013ca1d65c1ee7238d388d60bf4d08a86674cc64e857ef8d","tokens":15998}
9
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:lean:hilbert_functions_8","paper_id":"ai_hilbert_functions","token_sha256":"c90ec5d5c260eb681493db2f61d1fd883fc61f1810a7516e2a2191845714c148","tokens":13342}
10
+ {"id":"google-alphaproof-nexus:ai_hilbert_functions:nl:hilbert","paper_id":"ai_hilbert_functions","token_sha256":"40449098432cc6d134d42f71e839e9bd12baeedbd087ae7c5bdeef5285b420b0","tokens":4850}
11
+ {"id":"google-alphaproof-nexus:ai_optimization:lean:LastIterateConvergence","paper_id":"ai_optimization","token_sha256":"332827498046142840908e6acbb94bf415de0397e35a2552e26fcf9351cfcf01","tokens":15293}
12
+ {"id":"google-alphaproof-nexus:ai_quantumoptics:lean:35","paper_id":"ai_quantumoptics","token_sha256":"dba15692ed71df1aa7668300ed150e8df42dc02e06615563521fb990e0de4be3","tokens":46294}
13
+ {"id":"google-alphaproof-nexus:ai_quantumoptics:lean:MonochromaticQuantumGraph","paper_id":"ai_quantumoptics","token_sha256":"cbac073a055a1f500e379cb3b9bc7ae4d98cd2df14221e3a826a84934efeccb6","tokens":12200}
14
+ {"id":"google-alphaproof-nexus:erdos_12:lean:erdos_12.parts.i","paper_id":"erdos_12","token_sha256":"0d3b09d15dfc61ef31f4e943a1908209fc435cf3afcba8114feab8b2d24bf6cc","tokens":23134}
15
+ {"id":"google-alphaproof-nexus:erdos_12:lean:erdos_12.parts.ii","paper_id":"erdos_12","token_sha256":"b962dbb2216df15f4e996720e8d2af43de52f00ed19010ea5701ae7b8441c356","tokens":17149}
16
+ {"id":"google-alphaproof-nexus:erdos_12:nl:erdos12i","paper_id":"erdos_12","token_sha256":"cd05bb9308f1174601a07eb60919dba9598759a157f0736057b4603b578ace1d","tokens":1599}
17
+ {"id":"google-alphaproof-nexus:erdos_12:nl:erdos12ii","paper_id":"erdos_12","token_sha256":"35a4bfc3b95227e248f4b36c125616acde6bf8ac05bc560100fb3223772d3def","tokens":1490}
18
+ {"id":"google-alphaproof-nexus:erdos_125:lean:erdos_125.variants.positive_lower_density","paper_id":"erdos_125","token_sha256":"1ad9c4e8cd87c56b73d01b39f2e2d677aec1f3add6c27017f5007ed21f966a8b","tokens":10719}
19
+ {"id":"google-alphaproof-nexus:erdos_125:nl:erdos125","paper_id":"erdos_125","token_sha256":"cd47420d21bf4e773592923561648f8bcb80912e11535b6f7c94e16752a23ab3","tokens":1084}
20
+ {"id":"google-alphaproof-nexus:erdos_138:lean:erdos_138.variants.difference","paper_id":"erdos_138","token_sha256":"48b467c05e0b4ed44928f696f48847739c2b13e4bf78831c4d9e9d8d4eb83e51","tokens":15380}
21
+ {"id":"google-alphaproof-nexus:erdos_138:nl:erdos138","paper_id":"erdos_138","token_sha256":"dd30bfe3bfbfeeb5a1c7429c86793f221e97146f5aad0d27bdd4aa19e4c26082","tokens":456}
22
+ {"id":"google-alphaproof-nexus:erdos_152:lean:erdos_152","paper_id":"erdos_152","token_sha256":"b1162400c0fca8c78c3648ed2e09e5e5f30ed4d29705b24aab7d13a2d7734fae","tokens":15025}
23
+ {"id":"google-alphaproof-nexus:erdos_152:nl:erdos152","paper_id":"erdos_152","token_sha256":"29950c7a7a84e1d7f961f7d5aed9775d5221e3971780e14bd78b1394bd9731d8","tokens":1389}
24
+ {"id":"google-alphaproof-nexus:erdos_26:lean:erdos_26.variants.tenenbaum","paper_id":"erdos_26","token_sha256":"231f4330a31798cdde29ad984a8e0f6204ff535a77ad83bd7f346eb0f10e9200","tokens":21364}
25
+ {"id":"google-alphaproof-nexus:erdos_26:nl:erdos26","paper_id":"erdos_26","token_sha256":"a9f1a0b0ed33dbc5740622189ccc4fbacccb82dd5a5cc9928db7cd1215aab940","tokens":1212}
26
+ {"id":"google-alphaproof-nexus:erdos_741:lean:erdos_741.parts.i","paper_id":"erdos_741","token_sha256":"a33eaa539e47f258a9e8b787aa592f00ec62fa6de519561ec4cd32e8d3a66055","tokens":41206}
27
+ {"id":"google-alphaproof-nexus:erdos_741:lean:erdos_741.parts.ii","paper_id":"erdos_741","token_sha256":"b2185c1097b68fe7e28117b480c6846b88b1799ad0ef38de4c266239da2b6bc3","tokens":10071}
28
+ {"id":"google-alphaproof-nexus:erdos_741:nl:erdos741i","paper_id":"erdos_741","token_sha256":"81742953a0671f75fe4235c0631e2aec5ce7b6a1d175ff2ab6b4d288387dd7ca","tokens":949}
29
+ {"id":"google-alphaproof-nexus:erdos_741:nl:erdos741ii","paper_id":"erdos_741","token_sha256":"e4d682a89d7d8a7d6933f07888c4e82c8b45299f3514904415512ae034713415","tokens":707}
30
+ {"id":"google-alphaproof-nexus:erdos_846:lean:erdos_846","paper_id":"erdos_846","token_sha256":"2bbc79040d0132f923901f3768e3419ef461dfa3fd0d7bf1ab67e28618601afd","tokens":22856}
31
+ {"id":"google-alphaproof-nexus:erdos_846:nl:erdos846","paper_id":"erdos_846","token_sha256":"989fc3f8af1c7a622d6f88da4c70b966adf195386e9b21d627da7db85f1c8631","tokens":1156}
32
+ {"id":"google-alphaproof-nexus:oeis_103311:lean:oeis_103311_conjecture_0","paper_id":"oeis_103311","token_sha256":"89b42085a0f51474e16964455ef5265deea441a7460d5f734a15cfc7855ea4c4","tokens":4467}
33
+ {"id":"google-alphaproof-nexus:oeis_108:lean:oeis_108_conjecture_2","paper_id":"oeis_108","token_sha256":"7ce51e40a3b61bdab121631b5cb29f5d600ae7f301be1fdf993ec8ee41634a6b","tokens":8129}
34
+ {"id":"google-alphaproof-nexus:oeis_113254:lean:oeis_113254_conjecture_0","paper_id":"oeis_113254","token_sha256":"a6cb3ed7dac482534fc0bbc7b3d1143c45965c5c9199c6752a4c9102b8a61287","tokens":1745}
35
+ {"id":"google-alphaproof-nexus:oeis_175386:lean:oeis_175386_conjecture_0","paper_id":"oeis_175386","token_sha256":"912d0b51cd9f8d3e5f514fb609cd70914d19efe805d4be195bafde371bd19714","tokens":8410}
36
+ {"id":"google-alphaproof-nexus:oeis_194806:lean:oeis_194806_conjecture_0","paper_id":"oeis_194806","token_sha256":"9120bf98a63d307f9947e2acaaaeb464eb3601e3dfd7d050d38f3b4d43aeaac8","tokens":9407}
37
+ {"id":"google-alphaproof-nexus:oeis_227582:lean:oeis_227582_conjecture_0","paper_id":"oeis_227582","token_sha256":"7b95a7f0c2f014062b4f47b9687e6858a792503b60601e357d0faf3389b6ee58","tokens":10559}
38
+ {"id":"google-alphaproof-nexus:oeis_228143:lean:oeis_228143_conjecture_1","paper_id":"oeis_228143","token_sha256":"bc6d83d149963d1b8e2c8a14a2e3936b45f2d8d59a05fa0bedb4ac2708162fc4","tokens":15198}
39
+ {"id":"google-alphaproof-nexus:oeis_248802:lean:oeis_248802_conjecture_0","paper_id":"oeis_248802","token_sha256":"4bcb832e1ff224c5a7f82aa5481227da922e5697e6e981565aaa53947d274b86","tokens":3437}
40
+ {"id":"google-alphaproof-nexus:oeis_248802:lean:oeis_248802_conjecture_4","paper_id":"oeis_248802","token_sha256":"365f51d6042193a445d2288ea6b96fd63927e3fdeb090eaf6d65abd286457f7c","tokens":15192}
41
+ {"id":"google-alphaproof-nexus:oeis_256012:lean:oeis_256012_conjecture_0","paper_id":"oeis_256012","token_sha256":"8a7d876769dd0e03c734f4fcaf98922f96ef994b2e4b15882c465d74c0168b0b","tokens":2184}
42
+ {"id":"google-alphaproof-nexus:oeis_267581:lean:oeis_267581_conjecture_0","paper_id":"oeis_267581","token_sha256":"ca1ad4b63fd81cb7e7798eb4ddd86cb50909a0f3aa6648997c01ab088df92171","tokens":3833}
43
+ {"id":"google-alphaproof-nexus:oeis_271591:lean:oeis_271591_conjecture_0","paper_id":"oeis_271591","token_sha256":"c59efe18ec2054171ea530c1326830ba9618e63e65626311a2b048fa46399513","tokens":10499}
44
+ {"id":"google-alphaproof-nexus:oeis_278070:lean:oeis_278070_conjecture_0","paper_id":"oeis_278070","token_sha256":"ed43bd44a31a45e3482e261647c7c19d6839b8ac4bc5c1abdc559bbce3de1885","tokens":1369}
45
+ {"id":"google-alphaproof-nexus:oeis_282779:lean:oeis_282779_conjecture_0","paper_id":"oeis_282779","token_sha256":"c90ac368bb226a1130493842c448550e8f803ff2c18b09ccbd12a82029d95753","tokens":3356}
46
+ {"id":"google-alphaproof-nexus:oeis_289411:lean:oeis_289411_conjecture_0","paper_id":"oeis_289411","token_sha256":"f48b1fd184a61844250f98ec5b9e30115ebe1eed19f71baa4745c229f6086185","tokens":3477}
47
+ {"id":"google-alphaproof-nexus:oeis_2897:lean:oeis_2897_conjecture_0","paper_id":"oeis_2897","token_sha256":"e329874eae16693f6ad8df015ea4ff8b4a0118908a157e519408fa2c3ebc931e","tokens":14118}
48
+ {"id":"google-alphaproof-nexus:oeis_306424:lean:oeis_306424_conjecture_0","paper_id":"oeis_306424","token_sha256":"8a3819d24954886ef594f45651cc5da5516c67a2626a2d15a588ab53d843efdc","tokens":6997}
49
+ {"id":"google-alphaproof-nexus:oeis_307865:lean:oeis_307865_conjecture_0","paper_id":"oeis_307865","token_sha256":"d3b8b71bc43649920c3d12028f54e5c096b6bae0eb0f3b2ebecc3505ecd8f821","tokens":4006}
50
+ {"id":"google-alphaproof-nexus:oeis_323557:lean:oeis_323557_conjecture_0","paper_id":"oeis_323557","token_sha256":"7b0b3a3cf80b233cc445bf338ab0ea409d940899867c9ea87b4a61962e205698","tokens":5080}
51
+ {"id":"google-alphaproof-nexus:oeis_340737:lean:oeis_340737_conjecture_0","paper_id":"oeis_340737","token_sha256":"22d18094fe30efbe1c22ef340ccfa053b284f29bb6e0db3fd0caa41177a374cf","tokens":10844}
52
+ {"id":"google-alphaproof-nexus:oeis_341254:lean:oeis_341254_conjecture_0","paper_id":"oeis_341254","token_sha256":"322ebdb82a83ec38fda3a98950e93797da502c5c3de295414eca95a706d5050f","tokens":3195}
53
+ {"id":"google-alphaproof-nexus:oeis_363347:lean:oeis_363347_conjecture_2","paper_id":"oeis_363347","token_sha256":"bf6ba557044c7a77aeb1cedc0f2672c72657a4d85d700f929941f0b598a32ab2","tokens":19330}
54
+ {"id":"google-alphaproof-nexus:oeis_372761:lean:oeis_372761_conjecture_2","paper_id":"oeis_372761","token_sha256":"fe3eb812e1bbb186497a46a76efc515b26b36949d431af48963a322f3398eb7d","tokens":18836}
55
+ {"id":"google-alphaproof-nexus:oeis_51293:lean:oeis_51293_conjecture_0","paper_id":"oeis_51293","token_sha256":"16a272d1832dc8d315349381a1de2c5a847a54feb93fc8dc1efa973dd74266a3","tokens":19789}
56
+ {"id":"google-alphaproof-nexus:oeis_51293:nl:oeis_51293","paper_id":"oeis_51293","token_sha256":"93118b5673f21dbe41116a2c40b90141f3709924fc531280ad7545c82b0a5f72","tokens":1797}
57
+ {"id":"google-alphaproof-nexus:oeis_62567:lean:oeis_62567_conjecture_0","paper_id":"oeis_62567","token_sha256":"bc17230c27885a430d770757fd4ab13f816ac2a0d3114015f7b67ef0875ee4b8","tokens":7096}
58
+ {"id":"google-alphaproof-nexus:oeis_a028859:lean:oeis_A028859_conjecture_1","paper_id":"oeis_a028859","token_sha256":"161b1da9711e062d31d5dd712d4f1a57827309790369d65ef8d62916d938d8dd","tokens":9895}
59
+ {"id":"google-alphaproof-nexus:oeis_a091669:lean:a091669_conjecture_primitive_root","paper_id":"oeis_a091669","token_sha256":"f88e6e0331996a10b7fcb2e9f62c7f071e38ef781b26b37177373228f5689f7e","tokens":9479}
60
+ {"id":"google-alphaproof-nexus:oeis_a211417:lean:oeis_a211417_conjecture_specific","paper_id":"oeis_a211417","token_sha256":"7049c73fefdd4258a0f6e14e63fa14c08c66ba05f80623a5b68b4bc0b25f4aa0","tokens":6172}
61
+ {"id":"google-alphaproof-nexus:oeis_a224515:lean:A224515_conjecture_existence","paper_id":"oeis_a224515","token_sha256":"910dd83b592e059e95dc38c8c2a639c93bb7d43fa61cbbbbfae642dbeb86c3a2","tokens":5499}
62
+ {"id":"google-alphaproof-nexus:oeis_a237271:lean:oeis_a237271_conjecture_2","paper_id":"oeis_a237271","token_sha256":"2ca8dfd80be72fc377e8b49e673c56f5eea0516dec3e4963df575ca21eb55e54","tokens":1757}
63
+ {"id":"google-alphaproof-nexus:oeis_a258667:lean:oeis_A258667_conjecture_0","paper_id":"oeis_a258667","token_sha256":"a50eb2edcadafcd4b38b0d0ca755f406ebe1b4c55cc770d2412adc2263efdd94","tokens":13029}
64
+ {"id":"google-alphaproof-nexus:oeis_a300997:lean:oeis_a300997_finite_difference_is_one_or_two","paper_id":"oeis_a300997","token_sha256":"a1c66ef6fc30e486971e0245cb3001dfaaa642e094ce1cf8d27109e5aa544477","tokens":19198}
65
+ {"id":"google-alphaproof-nexus:oeis_a309132:lean:A309132_conjecture_carmichael","paper_id":"oeis_a309132","token_sha256":"d5287bd7fbdf4bfad357ea5a07d71c4eadac0ad66d1fc3ceae3cf596623e7fb9","tokens":12965}
66
+ {"id":"google-alphaproof-nexus:oeis_a325046:lean:a325046_odd_terms_at_k_times_k_plus_1","paper_id":"oeis_a325046","token_sha256":"4bb30e8baf35a65951bdd90263fc78ea5fdbfb8708951f092f82520549598882","tokens":4848}
67
+ {"id":"google-alphaproof-nexus:oeis_a363102:lean:oeis_a363102_conjecture_1","paper_id":"oeis_a363102","token_sha256":"3b127a45f7e60c7ae407cb6073f469b9deab842ef4dddfa310204bf6da48dda8","tokens":7883}
68
+ {"id":"google-alphaproof-nexus:oeis_a368692:lean:oeis_a368692_conjecture_integrality","paper_id":"oeis_a368692","token_sha256":"2a7c78cd64707e5862f13a46cdd06a32ee71d92596ecde6bb6707528c9809ce9","tokens":4307}
69
+ {"id":"google-alphaproof-nexus:oeis_a382590:lean:A382590_conjecture_kth_prime_factor_is_eventually_periodic","paper_id":"oeis_a382590","token_sha256":"636fbcc573bc988be4d40462440dfb6f8244c0bc890700c22bcb162a7bccaded","tokens":4984}
70
+ {"id":"google-alphaproof-nexus:stacks_project:lean:AlgebrasOverFields_GeometricallyIrreducibleAlgebra_of_directed_colimit","paper_id":"stacks_project","token_sha256":"c0a5dbb3e04aa6058107707a8faa1a5e641fa9333dd883f1f5f5aa1783d02c08","tokens":17362}
71
+ {"id":"google-alphaproof-nexus:stacks_project:lean:AlgebrasOverFields_GeometricallyIrreducibleAlgebra_transitive","paper_id":"stacks_project","token_sha256":"a51e272ae99a39f0871f9181f3f1f172994755b8145b9348112020afbafda8ac","tokens":9010}
72
+ {"id":"google-alphaproof-nexus:stacks_project:lean:AlgebrasOverFields_geometrically_irreducible_tfae","paper_id":"stacks_project","token_sha256":"838db9368b500f34f9636ab3b618ebb0540c13b222a7fc50a920812b809507a5","tokens":28240}
73
+ {"id":"google-alphaproof-nexus:stacks_project:lean:AlgebrasOverFields_minimalPrimes_bijection_of_geometrically_IrreducibleAlgebra","paper_id":"stacks_project","token_sha256":"0b1b1fcc6fce80b4901b3545a54b3d74dd66a5acb009b1677d706fb44bbbdaeb","tokens":11990}
74
+ {"id":"google-alphaproof-nexus:stacks_project:lean:AlgebrasOverFields_unique_minimal_prime_of_tensor_product_of_separably_closed","paper_id":"stacks_project","token_sha256":"be84fc5c6ef47fc8801a56749aaadbc3f88933f9bf6b986db2b0fbe1d0f41063","tokens":18383}
75
+ {"id":"google-alphaproof-nexus:stacks_project:lean:GeometricallyReduced_isGeometricallyReduced_iff_of_separable_algebraic","paper_id":"stacks_project","token_sha256":"8c66a9560cc62c1f128c4c5d378fa6e06618e9496fdf5f7d9eaa501d4c777181","tokens":23720}
76
+ {"id":"google-alphaproof-nexus:stacks_project:lean:GeometricallyReduced_isReduced_tensorProduct_of_separable","paper_id":"stacks_project","token_sha256":"fbf40b657061172e5ca92178bc4cb83023137fc97ad30b72f981b5414996d3f9","tokens":10167}
77
+ {"id":"google-alphaproof-nexus:stacks_project:lean:GeometricallyReduced_isReduced_tensorProduct_of_separablyGenerated","paper_id":"stacks_project","token_sha256":"196c4382faed3558a3b6fad983a1179519b63baac8cc56e458722752f57fec02","tokens":21676}
78
+ {"id":"google-alphaproof-nexus:stacks_project:lean:NormalRings_NormalDomain_powerSeries","paper_id":"stacks_project","token_sha256":"70e6ef8da9690110ab9b3a7e615397ecfab92b883b28c40d8864ff394e501840","tokens":12158}
79
+ {"id":"google-alphaproof-nexus:stacks_project:lean:NormalRings_normalRing_characterization","paper_id":"stacks_project","token_sha256":"7220f58551821098a8d3870b36293ec822f4280ec60e4ab148ce6b015a39b98f","tokens":23027}
80
+ {"id":"google-alphaproof-nexus:stacks_project:lean:Separability_exists_isPurelyInseparable_isSeparablyGenerated_of_essFiniteType","paper_id":"stacks_project","token_sha256":"67c9704625629723fa6e6a3b8c8d67983542f2d8c0d85ad5c47a001182a8bca0","tokens":41743}
components/google-alphaproof-nexus/packed/train-rows.jsonl ADDED
The diff for this file is too large to render. See raw diff
 
components/google-alphaproof-nexus/sources.json ADDED
@@ -0,0 +1,28 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "component": "google-alphaproof-nexus",
3
+ "upstream": {
4
+ "name": "AlphaProof Nexus results",
5
+ "repository": "https://github.com/google-deepmind/alphaproof-nexus-results",
6
+ "commit": "0647711a71183c1ea492ad60860776617ce1ea88",
7
+ "fixed_revision_url": "https://github.com/google-deepmind/alphaproof-nexus-results/tree/0647711a71183c1ea492ad60860776617ce1ea88",
8
+ "license": "Apache-2.0",
9
+ "license_file": "LICENSE",
10
+ "creators": [
11
+ "Google DeepMind"
12
+ ]
13
+ },
14
+ "changes": [
15
+ "Verbatim text projection of every upstream .lean file, Apache-2.0 copyright header retained, with no rewriting",
16
+ "pypdf text-layer extraction of every upstream natural-language proof PDF, with a lossy-extraction flag and per-PDF audit",
17
+ "Problem-stem grouping so a Lean proof and its prose proof never cross splits; directory grouping where no filename pairing exists",
18
+ "Exact text deduplication and deterministic group-level split assignment",
19
+ "Lossless JSONL/Parquet export; SentencePiece causal-window packing in the existing components/openai-math format"
20
+ ],
21
+ "all_records_license": "Apache-2.0",
22
+ "record_changes": [
23
+ "No text was generated, summarized, translated, or repaired",
24
+ "PDF text control characters were replaced and blank-line runs collapsed; the original bytes remain identified by sha256",
25
+ "No Lean compilation, proof checking, or mathematical verification was performed"
26
+ ],
27
+ "mathematical_verification": "not_performed"
28
+ }
components/google-alphaproof-nexus/statistics.json ADDED
@@ -0,0 +1,64 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "source_commit": "0647711a71183c1ea492ad60860776617ce1ea88",
3
+ "upstream_files": 92,
4
+ "lean_files": 71,
5
+ "pdf_files": 12,
6
+ "problem_groups": 51,
7
+ "records_before_deduplication": 83,
8
+ "records": 83,
9
+ "excluded_exact_duplicates": 0,
10
+ "record_types": {
11
+ "formal_lean_proof": 71,
12
+ "natural_language_proof": 12
13
+ },
14
+ "splits": {
15
+ "train": {
16
+ "records": 80,
17
+ "problems": 49,
18
+ "text_characters": 2154974,
19
+ "record_types": {
20
+ "formal_lean_proof": 69,
21
+ "natural_language_proof": 11
22
+ }
23
+ },
24
+ "dev": {
25
+ "records": 1,
26
+ "problems": 1,
27
+ "text_characters": 6805,
28
+ "record_types": {
29
+ "formal_lean_proof": 1
30
+ }
31
+ },
32
+ "test": {
33
+ "records": 2,
34
+ "problems": 1,
35
+ "text_characters": 67992,
36
+ "record_types": {
37
+ "formal_lean_proof": 1,
38
+ "natural_language_proof": 1
39
+ }
40
+ }
41
+ },
42
+ "pdf_extraction": {
43
+ "pages": 30,
44
+ "characters": 52415
45
+ },
46
+ "packed": {
47
+ "train": {
48
+ "rows": 491,
49
+ "tokens": 1004469,
50
+ "loss_bearing_labels": 1004389
51
+ },
52
+ "dev": {
53
+ "rows": 2,
54
+ "tokens": 3144,
55
+ "loss_bearing_labels": 3143
56
+ },
57
+ "test": {
58
+ "rows": 14,
59
+ "tokens": 28411,
60
+ "loss_bearing_labels": 28409
61
+ }
62
+ },
63
+ "mathematical_verification": "not_performed"
64
+ }
components/google-alphaproof-nexus/tokenizer/tokenizer.model ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18b5aa749c6c791ab59c2fd1eb39dfbaa155247322119d3924d3cb9768622417
3
+ size 503703
components/google-alphaproof-nexus/tools/build_google_alphaproof_nexus.py ADDED
@@ -0,0 +1,588 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+ """Build the independently versioned Google AlphaProof Nexus contribution.
3
+
4
+ Mirrors components/openai-math: same record core fields, same Parquet
5
+ encoding, same mmla-type2-packed-v1 packing code and split policy.
6
+ """
7
+ from __future__ import annotations
8
+
9
+ import argparse
10
+ import gzip
11
+ import hashlib
12
+ import json
13
+ import math
14
+ import re
15
+ import shutil
16
+ import subprocess
17
+ import sys
18
+ from collections import Counter, deque
19
+ from pathlib import Path
20
+
21
+ COMPONENT = 'google-alphaproof-nexus'
22
+ UPSTREAM_REPO = 'https://github.com/google-deepmind/alphaproof-nexus-results'
23
+ COMMIT = '0647711a71183c1ea492ad60860776617ce1ea88'
24
+ LICENSE_ID = 'Apache-2.0'
25
+ CORE = ('id', 'text', 'group', 'license', 'paper_id', 'split', 'record_type', 'source_url')
26
+ SPLITS = ('train', 'dev', 'test')
27
+ SEQ_LEN = 2048
28
+ SHARD_ROWS = 256
29
+ IGNORED_LOCAL = {'manifest.json', 'validation.json', '.DS_Store'}
30
+
31
+
32
+ def canonical(value):
33
+ return json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(',', ':'))
34
+
35
+
36
+ def sha(path):
37
+ h = hashlib.sha256()
38
+ with Path(path).open('rb') as f:
39
+ for chunk in iter(lambda: f.read(1024 * 1024), b''):
40
+ h.update(chunk)
41
+ return h.hexdigest()
42
+
43
+
44
+ def write_json(path, obj):
45
+ Path(path).parent.mkdir(parents=True, exist_ok=True)
46
+ Path(path).write_text(json.dumps(obj, ensure_ascii=False, indent=2) + '\n', encoding='utf-8')
47
+
48
+
49
+ def write_jsonl(path, rows):
50
+ path = Path(path)
51
+ path.parent.mkdir(parents=True, exist_ok=True)
52
+ if path.suffix == '.gz':
53
+ with path.open('wb') as raw, gzip.GzipFile(filename='', mode='wb', fileobj=raw, mtime=0) as gz:
54
+ for row in rows:
55
+ gz.write((canonical(row) + '\n').encode('utf-8'))
56
+ else:
57
+ with path.open('w', encoding='utf-8') as f:
58
+ for row in rows:
59
+ f.write(canonical(row) + '\n')
60
+
61
+
62
+ def write_parquet(path, rows):
63
+ import pyarrow as pa
64
+ import pyarrow.parquet as pq
65
+ columns = {key: [str(row.get(key) or '') for row in rows] for key in CORE}
66
+ columns['metadata_json'] = [canonical({k: v for k, v in row.items() if k not in CORE}) for row in rows]
67
+ schema = pa.schema([(key, pa.string()) for key in (*CORE, 'metadata_json')])
68
+ pq.write_table(pa.Table.from_pydict(columns, schema=schema), path, compression='zstd', row_group_size=1024)
69
+
70
+
71
+ def split_groups(groups):
72
+ """Identical policy to components/openai-math/tools/finalize_openai_math.py."""
73
+ ordered = sorted(groups, key=lambda g: (hashlib.sha256(g.encode()).hexdigest(), g))
74
+ if len(ordered) < 3:
75
+ raise ValueError('Need at least 3 problem groups for nonempty leakage-safe splits')
76
+ holdout = max(1, math.ceil(len(ordered) * .01))
77
+ if 2 * holdout >= len(ordered):
78
+ holdout = 1
79
+ return {g: ('dev' if i < holdout else 'test' if i < 2 * holdout else 'train') for i, g in enumerate(ordered)}
80
+
81
+
82
+ def pack_split(root, split, records, sp, seq_len=SEQ_LEN, shard_rows=SHARD_ROWS):
83
+ """Copied verbatim in behaviour from components/openai-math/tools/finalize_openai_math.py."""
84
+ import numpy as np
85
+ packed = root / 'packed'
86
+ packed.mkdir(exist_ok=True)
87
+ shard_entries, pending_rows, pending_labels, pending_attrs = [], [], [], []
88
+ buffer, docs, stream_start, total_tokens, label_count, row_number = [], deque(), 0, 0, 0, 0
89
+ attr_file = (packed / f'{split}-rows.jsonl').open('w', encoding='utf-8')
90
+ token_file = (packed / f'{split}-records.jsonl').open('w', encoding='utf-8')
91
+
92
+ def flush():
93
+ nonlocal pending_rows, pending_labels, pending_attrs
94
+ if not pending_rows:
95
+ return
96
+ shard = len(shard_entries)
97
+ fields = {}
98
+ arrays = {'input_ids': np.asarray(pending_rows, dtype=np.int32),
99
+ 'labels': np.asarray(pending_labels, dtype=np.int32),
100
+ 'role_ids': np.zeros((len(pending_rows), seq_len), dtype=np.int32)}
101
+ for name, array in arrays.items():
102
+ path = packed / f'{split}-{shard:06d}-{name}.npy'
103
+ np.save(path, array, allow_pickle=False)
104
+ fields[name] = {'path': path.name, 'sha256': sha(path), 'bytes': path.stat().st_size,
105
+ 'dtype': 'int32', 'shape': list(array.shape)}
106
+ shard_entries.append({'count': len(pending_rows), 'fields': fields})
107
+ for index, attr in enumerate(pending_attrs):
108
+ attr.update(shard=shard, shard_row=index)
109
+ attr_file.write(canonical(attr) + '\n')
110
+ pending_rows, pending_labels, pending_attrs = [], [], []
111
+
112
+ def emit():
113
+ nonlocal buffer, stream_start, label_count, row_number
114
+ n = min(seq_len, len(buffer))
115
+ values = buffer[:n]
116
+ targets = buffer[1:min(n + 1, len(buffer))]
117
+ if len(targets) < n:
118
+ targets.append(-100)
119
+ targets = [-100 if t == 1 else t for t in targets]
120
+ label_count += sum(t != -100 for t in targets)
121
+ pending_rows.append(values + [0] * (seq_len - n))
122
+ pending_labels.append(targets + [-100] * (seq_len - n))
123
+ stop = stream_start + n
124
+ spans = []
125
+ for doc in docs:
126
+ lo, hi = max(stream_start, doc['start']), min(stop, doc['end'])
127
+ if lo < hi:
128
+ spans.append({'record_id': doc['id'], 'paper_id': doc['paper_id'],
129
+ 'row_start': lo - stream_start, 'row_end': hi - stream_start,
130
+ 'record_start': lo - doc['start'], 'record_end': hi - doc['start']})
131
+ pending_attrs.append({'row': row_number, 'stream_start': stream_start, 'length': n, 'segments': spans})
132
+ row_number += 1
133
+ stream_start = stop
134
+ del buffer[:n]
135
+ while docs and docs[0]['end'] <= stream_start:
136
+ docs.popleft()
137
+ if len(pending_rows) == shard_rows:
138
+ flush()
139
+
140
+ for record in records:
141
+ tokens = [1] + sp.encode(record['text'], out_type=int) + [2]
142
+ array = np.asarray(tokens, dtype='<i4')
143
+ token_file.write(canonical({'id': record['id'], 'paper_id': record['paper_id'], 'tokens': len(tokens),
144
+ 'token_sha256': hashlib.sha256(array.tobytes()).hexdigest()}) + '\n')
145
+ docs.append({'id': record['id'], 'paper_id': record['paper_id'], 'start': total_tokens,
146
+ 'end': total_tokens + len(tokens)})
147
+ total_tokens += len(tokens)
148
+ buffer.extend(tokens)
149
+ while len(buffer) >= seq_len + 1:
150
+ emit()
151
+ if buffer:
152
+ emit()
153
+ flush()
154
+ attr_file.close()
155
+ token_file.close()
156
+ return shard_entries, {'rows': row_number, 'tokens': total_tokens, 'loss_bearing_labels': label_count}
157
+
158
+
159
+ def write_manifest(root):
160
+ entries = []
161
+ for path in sorted(root.rglob('*')):
162
+ relative = path.relative_to(root).as_posix()
163
+ if path.is_file() and relative not in IGNORED_LOCAL and '.tmp' not in path.name:
164
+ entries.append({'path': relative, 'bytes': path.stat().st_size, 'sha256': sha(path)})
165
+ write_json(root / 'manifest.json', {'format': 'cayley-contribution-file-manifest-v1',
166
+ 'excludes': sorted(IGNORED_LOCAL), 'files': entries})
167
+
168
+
169
+ def lean_group(rel):
170
+ """Problem-stem group for an APNOutputs Lean file."""
171
+ parts = Path(rel).parts
172
+ category, stem = parts[1], Path(rel).stem
173
+ if category == 'ErdosProblems':
174
+ return 'erdos_' + stem.split('_')[1].split('.')[0], 'erdos problem number from the filename stem'
175
+ if category == 'OEIS':
176
+ tokens = stem.split('_')
177
+ ident = tokens[1] if tokens[0].lower() == 'oeis' else tokens[0]
178
+ return 'oeis_' + ident.lower(), 'OEIS sequence identifier from the filename stem'
179
+ if category == 'StacksProject':
180
+ return 'stacks_project', 'no filename pairing; grouped by the containing problem directory APNOutputs/StacksProject'
181
+ if category == 'AICollaborator':
182
+ subject = parts[2]
183
+ if stem.startswith('hilbert_functions'):
184
+ return 'ai_hilbert_functions', 'paired with NaturalLanguageProofs/AICollaborator/hilbert.pdf by problem stem'
185
+ if stem == 'bipartite_graph_reconstruction_conjecture_2':
186
+ return 'ai_graph_reconstruction', 'paired with NaturalLanguageProofs/AICollaborator/graph_reconstruction.pdf by problem stem'
187
+ return 'ai_' + subject.lower(), f'no filename pairing; grouped by the containing problem directory APNOutputs/AICollaborator/{subject}'
188
+ raise ValueError(f'Unknown Lean category: {rel}')
189
+
190
+
191
+ def pdf_group(rel):
192
+ """Problem-stem group for a NaturalLanguageProofs PDF."""
193
+ category, stem = Path(rel).parts[1], Path(rel).stem
194
+ if category == 'ErdosProblems':
195
+ match = re.fullmatch(r'erdos(\d+)(i{1,2})?', stem)
196
+ if not match:
197
+ raise ValueError(f'Unparsed Erdos PDF name: {rel}')
198
+ return 'erdos_' + match.group(1), 'erdos problem number from the filename stem'
199
+ if category == 'OEIS':
200
+ match = re.fullmatch(r'oeis_(\d+)', stem.lower())
201
+ if not match:
202
+ raise ValueError(f'Unparsed OEIS PDF name: {rel}')
203
+ return 'oeis_' + match.group(1), 'OEIS sequence identifier from the filename stem'
204
+ if category == 'AICollaborator':
205
+ if stem == 'hilbert':
206
+ return 'ai_hilbert_functions', 'paired with APNOutputs/AICollaborator/AlgebraicGeometry/hilbert_functions_*.lean'
207
+ if stem == 'graph_reconstruction':
208
+ return 'ai_graph_reconstruction', 'paired with APNOutputs/AICollaborator/Graphs/bipartite_graph_reconstruction_conjecture_2.lean'
209
+ raise ValueError(f'Unpaired PDF name: {rel}')
210
+
211
+
212
+ PDF_ARTIFACT_RANGES = (
213
+ ('c0_c1_control', 0x00, 0x1F),
214
+ ('c1_control', 0x7F, 0x9F),
215
+ ('replacement_char', 0xFFFD, 0xFFFD),
216
+ ('private_use', 0xE000, 0xF8FF),
217
+ ('ligature_block', 0xFB00, 0xFB06),
218
+ ('combining_mark', 0x0300, 0x036F),
219
+ )
220
+
221
+
222
+ def extract_pdf(path):
223
+ import pypdf
224
+ reader = pypdf.PdfReader(str(path))
225
+ pages = [page.extract_text() or '' for page in reader.pages]
226
+ raw = '\n'.join(pages)
227
+ control = sum(1 for c in raw if (ord(c) < 32 and c not in '\n\t') or 0x7F <= ord(c) <= 0x9F)
228
+ text = ''.join(' ' if (ord(c) < 32 and c not in '\n\t') or 0x7F <= ord(c) <= 0x9F else c for c in raw)
229
+ text = '\n'.join(line.rstrip() for line in text.split('\n'))
230
+ text = re.sub(r'\n{3,}', '\n\n', text).strip() + '\n'
231
+ non_ascii = sum(1 for c in text if ord(c) > 127)
232
+ artifacts = Counter()
233
+ for name, lo, hi in PDF_ARTIFACT_RANGES:
234
+ artifacts[name] = sum(1 for c in text if lo <= ord(c) <= hi)
235
+ return {
236
+ 'text': text,
237
+ 'pages': len(pages),
238
+ 'raw_characters': len(raw),
239
+ 'characters': len(text),
240
+ 'page_characters': [len(page) for page in pages],
241
+ 'non_ascii_characters': non_ascii,
242
+ 'non_ascii_ratio': round(non_ascii / len(text), 6) if text else 0.0,
243
+ 'control_characters_replaced': control,
244
+ 'artifact_characters': {k: v for k, v in artifacts.items() if v},
245
+ }
246
+
247
+
248
+ def main():
249
+ parser = argparse.ArgumentParser(description=__doc__)
250
+ parser.add_argument('--component', type=Path, default=Path(__file__).resolve().parents[1])
251
+ parser.add_argument('--source-root', type=Path, required=True)
252
+ parser.add_argument('--tokenizer', type=Path, default=Path('/Users/zxy/Cayley-Math/tokenizer/tokenizer.model'))
253
+ parser.add_argument('--pack', action='store_true')
254
+ args = parser.parse_args()
255
+ root = args.component.resolve()
256
+ source = args.source_root.resolve()
257
+
258
+ live = subprocess.check_output(['git', '-C', str(source), 'rev-parse', 'HEAD'], text=True).strip()
259
+ if live != COMMIT:
260
+ raise ValueError(f'Upstream checkout {live} differs from pinned {COMMIT}')
261
+
262
+ lean_files = sorted(p.relative_to(source).as_posix() for p in source.rglob('*.lean'))
263
+ pdf_files = sorted(p.relative_to(source).as_posix() for p in source.rglob('*.pdf'))
264
+ all_files = sorted(p.relative_to(source).as_posix() for p in source.rglob('*') if p.is_file() and '.git' not in p.parts)
265
+ if not lean_files or not pdf_files:
266
+ raise ValueError('Upstream snapshot lacks Lean or PDF files')
267
+
268
+ problems, assignments = {}, {}
269
+ for rel in lean_files:
270
+ group, rule = lean_group(rel)
271
+ problems.setdefault(group, []).append(rel)
272
+ assignments[rel] = {'group': group, 'rule': rule}
273
+ for rel in pdf_files:
274
+ group, rule = pdf_group(rel)
275
+ if group not in problems:
276
+ raise ValueError(f'PDF {rel} derived group {group} has no Lean companion')
277
+ assignments[rel] = {'group': group, 'rule': rule}
278
+ split_of = split_groups(sorted(problems))
279
+ empty = [s for s in SPLITS if s not in split_of.values()]
280
+ if empty:
281
+ raise ValueError(f'Empty splits: {empty}')
282
+
283
+ records, seen, duplicates = [], {}, []
284
+ for rel in lean_files + pdf_files:
285
+ path = source / rel
286
+ digest = sha(path)
287
+ group = assignments[rel]['group']
288
+ stem = Path(rel).stem
289
+ if rel.endswith('.lean'):
290
+ text = path.read_text(encoding='utf-8')
291
+ record_type, kind, origin = 'formal_lean_proof', 'lean', 'source_verbatim'
292
+ quality = 'verbatim_source_text'
293
+ transform = 'identity_verbatim_bytes'
294
+ metrics = {'source_characters': len(text)}
295
+ category = rel.split('/')[1]
296
+ else:
297
+ extracted = extract_pdf(path)
298
+ text = extracted.pop('text')
299
+ record_type, kind, origin = 'natural_language_proof', 'nl', 'source_pdf_extraction'
300
+ quality = 'lossy_pdf_text_layer'
301
+ transform = 'pypdf_page_text_control_chars_replaced_blank_runs_collapsed'
302
+ metrics = extracted
303
+ category = rel.split('/')[1]
304
+ if not text.strip():
305
+ raise ValueError(f'Empty text projection: {rel}')
306
+ record = {
307
+ 'id': f'{COMPONENT}:{group}:{kind}:{stem}',
308
+ 'text': text,
309
+ 'group': group,
310
+ 'license': LICENSE_ID,
311
+ 'paper_id': group,
312
+ 'split': split_of[group],
313
+ 'record_type': record_type,
314
+ 'source_url': f'https://github.com/google-deepmind/alphaproof-nexus-results/blob/{COMMIT}/{rel}',
315
+ 'source_commit': COMMIT,
316
+ 'source_path': rel,
317
+ 'source_sha256': digest,
318
+ 'source_bytes': path.stat().st_size,
319
+ 'text_sha256': hashlib.sha256(text.encode('utf-8')).hexdigest(),
320
+ 'content_origin': origin,
321
+ 'transform': transform,
322
+ 'extraction_quality': quality,
323
+ 'extraction_metrics': metrics,
324
+ 'upstream_category': category,
325
+ 'creators': ['Google DeepMind'],
326
+ 'verification_status': 'unverified_source_claim',
327
+ 'lean_verified': False,
328
+ 'mathematical_verification': 'not_performed',
329
+ }
330
+ if record['text_sha256'] in seen:
331
+ duplicates.append({'excluded_id': record['id'], 'kept_id': seen[record['text_sha256']],
332
+ 'group': group, 'source_path': rel,
333
+ 'text_sha256': record['text_sha256']})
334
+ continue
335
+ seen[record['text_sha256']] = record['id']
336
+ records.append(record)
337
+ ids = [r['id'] for r in records]
338
+ if len(set(ids)) != len(ids):
339
+ raise ValueError('Duplicate record IDs')
340
+ records.sort(key=lambda r: (r['group'], r['record_type'], r['source_path']))
341
+
342
+ split_records = {}
343
+ for split in SPLITS:
344
+ selected = [r for r in records if r['split'] == split]
345
+ if not selected:
346
+ raise ValueError(f'Empty {split} split')
347
+ write_jsonl(root / f'data/{split}.jsonl.gz', selected)
348
+ write_parquet(root / f'data/{split}.parquet', selected)
349
+ split_records[split] = selected
350
+
351
+ audit = root / 'audit'
352
+ audit.mkdir(parents=True, exist_ok=True)
353
+ write_jsonl(audit / 'source-files.jsonl', [
354
+ {'path': rel, 'sha256': sha(source / rel), 'bytes': (source / rel).stat().st_size,
355
+ 'source_url': f'{UPSTREAM_REPO}/blob/{COMMIT}/{rel}'} for rel in all_files])
356
+ write_jsonl(audit / 'extraction.jsonl', [
357
+ {'source_path': r['source_path'], **r['extraction_metrics'], 'extraction_quality': r['extraction_quality'],
358
+ 'source_url': r['source_url']} for r in records if r['record_type'] == 'natural_language_proof'])
359
+ write_jsonl(audit / 'excluded-duplicates.jsonl', duplicates)
360
+ unpaired = sorted({(assignments[rel]['group'], Path(rel).parent.as_posix())
361
+ for rel in lean_files if 'no filename pairing' in assignments[rel]['rule']})
362
+ write_json(audit / 'group-splits.json', {
363
+ 'format': 'cayley-group-splits-v1',
364
+ 'source_commit': COMMIT,
365
+ 'policy': 'sorted SHA256(group), first ceil(1%) groups to dev, next ceil(1%) to test, rest train',
366
+ 'splits': split_of,
367
+ 'rules': {g: sorted({assignments[r]['rule'] for r in problems[g]}) for g in sorted(problems)},
368
+ 'unpaired_directories': [{'group': g, 'directory': d} for g, d in unpaired],
369
+ 'assignments': assignments,
370
+ })
371
+
372
+ packed_stats = {}
373
+ if args.pack:
374
+ import sentencepiece as spm
375
+ sp = spm.SentencePieceProcessor(model_file=str(args.tokenizer))
376
+ if (sp.pad_id(), sp.bos_id(), sp.eos_id(), sp.unk_id(), sp.vocab_size()) != (0, 1, 2, 3, 32768):
377
+ raise ValueError('Unexpected tokenizer special IDs or vocabulary size')
378
+ tokenizer_dir = root / 'tokenizer'
379
+ tokenizer_dir.mkdir(exist_ok=True)
380
+ if args.tokenizer.resolve() != (tokenizer_dir / 'tokenizer.model').resolve():
381
+ shutil.copy2(args.tokenizer, tokenizer_dir / 'tokenizer.model')
382
+ packed = root / 'packed'
383
+ packed.mkdir(exist_ok=True)
384
+ for stale in packed.glob('*.npy'):
385
+ stale.unlink()
386
+ shards = {}
387
+ for split in SPLITS:
388
+ shards[split], packed_stats[split] = pack_split(root, split, split_records[split], sp)
389
+ write_json(packed / 'manifest.json', {
390
+ 'format': 'mmla-type2-packed-v1', 'phase': 'pretrain', 'layout': 'lm', 'seq_len': SEQ_LEN,
391
+ 'tokenizer': {'kind': 'sentencepiece', 'vocab_size': 32768, 'sha256': sha(args.tokenizer),
392
+ 'path': '../tokenizer/tokenizer.model',
393
+ 'ids': {'pad': 0, 'bos': 1, 'eos': 2, 'unk': 3, 'boundary': 4}},
394
+ 'quota_units': 'tokens', 'source': COMPONENT, 'splits': shards, 'statistics': packed_stats,
395
+ 'packing': {'stride': SEQ_LEN, 'bos_per_record': True, 'eos_per_record': True,
396
+ 'mask_target_bos': True, 'mask_final_target': True, 'loss_mask': -100,
397
+ 'row_attribution': '{split}-rows.jsonl', 'record_tokens': '{split}-records.jsonl'},
398
+ })
399
+ write_json(audit / 'packing.json', {
400
+ 'packer': 'tools/build_google_alphaproof_nexus.py:pack_split (copied from components/openai-math)',
401
+ 'tokenizer_sha256': sha(tokenizer_dir / 'tokenizer.model'),
402
+ 'tokenizer_note': 'SentencePiece decode is whitespace-normalizing for runs of spaces; every packed row is '
403
+ 'verified by re-encoding and by decoded-text comparison in the spot check.',
404
+ 'sequence_length': SEQ_LEN, 'shard_rows': SHARD_ROWS,
405
+ })
406
+
407
+ license_source = source / 'LICENSE'
408
+ if not license_source.is_file():
409
+ raise ValueError('Upstream LICENSE missing')
410
+ shutil.copy2(license_source, root / 'LICENSE')
411
+
412
+ stats = {
413
+ 'source_commit': COMMIT,
414
+ 'upstream_files': len(all_files),
415
+ 'lean_files': len(lean_files),
416
+ 'pdf_files': len(pdf_files),
417
+ 'problem_groups': len(problems),
418
+ 'records_before_deduplication': len(records) + len(duplicates),
419
+ 'records': len(records),
420
+ 'excluded_exact_duplicates': len(duplicates),
421
+ 'record_types': dict(Counter(r['record_type'] for r in records)),
422
+ 'splits': {s: {'records': len(v), 'problems': len({r['group'] for r in v}),
423
+ 'text_characters': sum(len(r['text']) for r in v),
424
+ 'record_types': dict(Counter(r['record_type'] for r in v))} for s, v in split_records.items()},
425
+ 'pdf_extraction': {'pages': sum(r['extraction_metrics']['pages'] for r in records
426
+ if r['record_type'] == 'natural_language_proof'),
427
+ 'characters': sum(r['extraction_metrics']['characters'] for r in records
428
+ if r['record_type'] == 'natural_language_proof')},
429
+ 'packed': packed_stats,
430
+ 'mathematical_verification': 'not_performed',
431
+ }
432
+ write_json(root / 'statistics.json', stats)
433
+ write_json(root / 'sources.json', {
434
+ 'component': COMPONENT,
435
+ 'upstream': {
436
+ 'name': 'AlphaProof Nexus results',
437
+ 'repository': UPSTREAM_REPO,
438
+ 'commit': COMMIT,
439
+ 'fixed_revision_url': f'{UPSTREAM_REPO}/tree/{COMMIT}',
440
+ 'license': LICENSE_ID,
441
+ 'license_file': 'LICENSE',
442
+ 'creators': ['Google DeepMind'],
443
+ },
444
+ 'changes': [
445
+ 'Verbatim text projection of every upstream .lean file, Apache-2.0 copyright header retained, with no rewriting',
446
+ 'pypdf text-layer extraction of every upstream natural-language proof PDF, with a lossy-extraction flag and per-PDF audit',
447
+ 'Problem-stem grouping so a Lean proof and its prose proof never cross splits; directory grouping where no filename pairing exists',
448
+ 'Exact text deduplication and deterministic group-level split assignment',
449
+ 'Lossless JSONL/Parquet export; SentencePiece causal-window packing in the existing components/openai-math format',
450
+ ],
451
+ 'all_records_license': LICENSE_ID,
452
+ 'record_changes': [
453
+ 'No text was generated, summarized, translated, or repaired',
454
+ 'PDF text control characters were replaced and blank-line runs collapsed; the original bytes remain identified by sha256',
455
+ 'No Lean compilation, proof checking, or mathematical verification was performed',
456
+ ],
457
+ 'mathematical_verification': 'not_performed',
458
+ })
459
+
460
+ readme = f"""# Google AlphaProof Nexus contribution to Cayley-Math
461
+
462
+ This independently versioned component contains source-derived mathematical text from Google DeepMind's
463
+ AlphaProof Nexus result repository at commit {COMMIT}. It follows Cayley-Math's components/ convention and
464
+ reuses the record schema, Parquet encoding, split policy and packed-array format of
465
+ `components/openai-math/`. The frozen Phase 1 mirror is unchanged. Upstream works are attributed to Google
466
+ DeepMind and distributed under Apache-2.0; see LICENSE, sources.json, and audit/.
467
+
468
+ ## What this is, and what it is not
469
+
470
+ These are proofs of **specific solved problems**, not a general corpus or a proof-search benchmark. Upstream
471
+ publishes only the problems where AlphaProof Nexus found a successful proof, so the collection is
472
+ selection-biased and small. Every upstream file carries Google's Apache-2.0 copyright header; the Lean files
473
+ retain it verbatim.
474
+
475
+ No Lean compilation, `lake build`, Comparator run, or mathematical verification was performed. The component
476
+ does not establish mathematical correctness of any statement, formalization, or proof. Upstream states that
477
+ the Lean outputs were mechanically formalized, but this contribution only transports text.
478
+
479
+ ## Coverage
480
+
481
+ The pinned snapshot has **{len(all_files)} files**, including **{len(lean_files)} `.lean` files** and
482
+ **{len(pdf_files)} natural-language proof PDFs**. Projection yields **{len(records)} records** across
483
+ **{len(problems)} problem groups**: **{stats['record_types'].get('formal_lean_proof', 0)}** `formal_lean_proof`
484
+ and **{stats['record_types'].get('natural_language_proof', 0)}** `natural_language_proof`. Exact text
485
+ deduplication removed {len(duplicates)} records; audit/excluded-duplicates.jsonl lists them.
486
+
487
+ Splits are assigned per problem group so a Lean proof and its prose proof never cross train/dev/test.
488
+ Assignment sorts SHA256(group), reserves the first ceil(1%) of groups for development, the next ceil(1%) for
489
+ test, and the rest for training. Assigned counts: train {len(split_records['train'])} records over
490
+ {len({r['group'] for r in split_records['train']})} groups, dev {len(split_records['dev'])} over
491
+ {len({r['group'] for r in split_records['dev']})}, test {len(split_records['test'])} over
492
+ {len({r['group'] for r in split_records['test']})}.
493
+
494
+ Grouping derives from the problem stem, for example `erdos_152` from `erdos_152.lean` plus `erdos152.pdf`,
495
+ `oeis_51293`, or `ai_hilbert_functions` from `hilbert_functions_*.lean` plus `hilbert.pdf`. Where no filename
496
+ pairing exists, the containing problem directory is the group; audit/group-splits.json records each
497
+ assignment rule and the unpaired directories. Stacks Project files are flat in one directory and therefore
498
+ share the single group `stacks_project`.
499
+
500
+ ## Text projection
501
+
502
+ `formal_lean_proof` records are the upstream file text, byte-identical to the pinned blob, license header
503
+ included. `natural_language_proof` records come from pypdf page-text extraction of each PDF. Extraction is
504
+ lossy: math symbols arrive as Unicode substitutions, ligature and encoding artifacts appear (for example
505
+ `Erd˝os`), and inline formulas lose their two-dimensional layout. C0/C1 control characters were replaced
506
+ and blank-line runs collapsed; nothing else was altered. Each PDF record carries `extraction_quality =
507
+ lossy_pdf_text_layer` plus page count, character counts, non-ASCII ratio and artifact counts in
508
+ `extraction_metrics`; audit/extraction.jsonl holds the same per-PDF audit. Treat prose records as noisy text,
509
+ not as faithful renderings of the PDFs.
510
+
511
+ ## Files and schema
512
+
513
+ - data/{{train,dev,test}}.jsonl.gz: source records with required fields text, group, license; unique id, paper_id, split, record_type, source_url, and provenance.
514
+ - 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 serialized in metadata_json. Merge json.loads(metadata_json) with the core columns to reconstruct each JSONL record.
515
+ - audit/source-files.jsonl: path, sha256, bytes and pinned source_url for every upstream blob.
516
+ - audit/group-splits.json: group assignments, per-group rule and unpaired directories.
517
+ - audit/extraction.jsonl: per-PDF extraction metrics and quality flag.
518
+ - audit/excluded-duplicates.jsonl: exact-text duplicates removed before packaging.
519
+ - statistics.json: measured record, group, split and packed-token counts.
520
+ - manifest.json: SHA256 and byte size for delivered files except itself and validation.json.
521
+ - packed/: mmla-type2-packed-v1 arrays, manifest, row attribution and per-record token counts.
522
+ - tools/: the build and validation scripts used to produce and check these files.
523
+
524
+ ## Loading
525
+
526
+ ~~~python
527
+ import gzip, json
528
+ from pathlib import Path
529
+ component = Path('components/google-alphaproof-nexus')
530
+ with gzip.open(component / 'data/train.jsonl.gz', 'rt', encoding='utf-8') as f:
531
+ first_record = json.loads(next(f))
532
+
533
+ import pyarrow.parquet as pq
534
+ row = pq.read_table(component / 'data/train.parquet').slice(0, 1).to_pylist()[0]
535
+ metadata = json.loads(row.pop('metadata_json'))
536
+ first_parquet_record = {{**row, **metadata}}
537
+ ~~~
538
+
539
+ ## Packed training arrays
540
+
541
+ packed/manifest.json uses `mmla-type2-packed-v1`: input_ids, labels and role_ids are int32 arrays of shape
542
+ [rows, {SEQ_LEN}], pad=0, role=0, ignored labels=-100. Each record is encoded as BOS + SentencePiece(text) +
543
+ EOS with the included Cayley-Math tokenizer (32768 pieces); window stride is {SEQ_LEN} with one-token lookahead,
544
+ labels predict the following token, and targets that cross into a new record's BOS or into padding are masked.
545
+ Split-specific rows.jsonl maps every non-padding token back to its record; records.jsonl stores token counts
546
+ and hashes. This component does not reproduce Phase 1 quotas, packing, or splits.
547
+
548
+ Measured packed counts: train {packed_stats.get('train', {}).get('rows', 0)} rows / {packed_stats.get('train', {}).get('tokens', 0)} tokens / {packed_stats.get('train', {}).get('loss_bearing_labels', 0)} loss-bearing labels; dev {packed_stats.get('dev', {}).get('rows', 0)} / {packed_stats.get('dev', {}).get('tokens', 0)} / {packed_stats.get('dev', {}).get('loss_bearing_labels', 0)}; test {packed_stats.get('test', {}).get('rows', 0)} / {packed_stats.get('test', {}).get('tokens', 0)} / {packed_stats.get('test', {}).get('loss_bearing_labels', 0)}.
549
+
550
+ ## Reproduction and verification
551
+
552
+ ~~~sh
553
+ python tools/build_google_alphaproof_nexus.py --component . --source-root /path/to/pinned/alphaproof-nexus-results --tokenizer tokenizer/tokenizer.model --pack
554
+ python tools/validate_google_alphaproof_nexus.py --component . --source-root /path/to/pinned/alphaproof-nexus-results
555
+ python tools/spotcheck_google_alphaproof_nexus.py --component . --write audit/spotcheck.json
556
+ ~~~
557
+
558
+ Dependencies: Python 3.13, numpy, sentencepiece, pyarrow, pypdf. The validator checks file hashes, source
559
+ coverage, exact-text uniqueness, group split separation, lossless Parquet reconstruction, packed array
560
+ shapes, causal shift, padding and boundary masks, token-level re-encoding of every record, and row
561
+ attribution. The spot check decodes packed rows back to text and round-trips a Parquet row to its JSONL
562
+ record. Neither script assesses theorem truth or proof completeness.
563
+
564
+ ## Limitations
565
+
566
+ - PDF extraction is lossy for symbols, ligatures, and formula layout; prose records are noisy.
567
+ - Only solved problems are present upstream, so the collection is not representative.
568
+ - No Lean compilation or proof verification; `lean-toolchain` pins `leanprover/lean4:v4.27.0` and
569
+ `formal_conjectures` is a required dependency, but this snapshot was not built.
570
+ - Near-duplicate detection, Phase 1 overlap analysis, and editorial review were not performed.
571
+ - The SentencePiece model cannot reproduce repeated runs of spaces: decoding drops them, so every packed row
572
+ re-encodes to the same token IDs while its decoded text equals the source only after collapsing whitespace
573
+ runs. This is inherited from the shared Cayley-Math tokenizer and applies to both record types.
574
+ - 12 records are prose and 1 test group (`ai_graph_reconstruction`) carries both record types, so the test
575
+ split holds 2 of the 83 records. Monitoring splits are small by construction at this corpus size.
576
+ """
577
+ (root / 'README.md').write_text(readme, encoding='utf-8')
578
+ spotcheck = root / 'tools/spotcheck_google_alphaproof_nexus.py'
579
+ if spotcheck.is_file():
580
+ subprocess.run([sys.executable, str(spotcheck), '--component', str(root),
581
+ '--write', 'audit/spotcheck.json'],
582
+ check=True, capture_output=True, text=True)
583
+ write_manifest(root)
584
+ print(json.dumps(stats, ensure_ascii=False, indent=2))
585
+
586
+
587
+ if __name__ == '__main__':
588
+ main()
components/google-alphaproof-nexus/tools/spotcheck_google_alphaproof_nexus.py ADDED
@@ -0,0 +1,233 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+ """Independent spot check: decode packed rows back to source text, round-trip Parquet, re-read splits."""
3
+ from __future__ import annotations
4
+ import argparse
5
+ import gzip
6
+ import hashlib
7
+ import json
8
+ import re
9
+ from pathlib import Path
10
+
11
+ SPLITS = ('train', 'dev', 'test')
12
+
13
+
14
+ def sha(path):
15
+ h = hashlib.sha256()
16
+ with Path(path).open('rb') as f:
17
+ for chunk in iter(lambda: f.read(1 << 20), b''):
18
+ h.update(chunk)
19
+ return h.hexdigest()
20
+
21
+
22
+ def canonical(value):
23
+ return json.dumps(value, ensure_ascii=False, sort_keys=True, separators=(',', ':'))
24
+
25
+
26
+ def read_jsonl(path):
27
+ opener = gzip.open if str(path).endswith('.gz') else open
28
+ with opener(path, 'rt', encoding='utf-8') as f:
29
+ return [json.loads(line) for line in f if line.strip()]
30
+
31
+
32
+ def normalize(text):
33
+ return re.sub(r'[ \t]+', ' ', re.sub(r'\n+', '\n', text)).strip()
34
+
35
+
36
+ def main():
37
+ parser = argparse.ArgumentParser(description=__doc__)
38
+ parser.add_argument('--component', type=Path, default=Path(__file__).resolve().parents[1])
39
+ parser.add_argument('--write', type=Path, default=Path('audit/spotcheck.json'))
40
+ args = parser.parse_args()
41
+ import numpy as np
42
+ import sentencepiece as spm
43
+ root = args.component.resolve()
44
+
45
+ tokenizer_path = root / 'tokenizer/tokenizer.model'
46
+ sp = spm.SentencePieceProcessor(model_file=str(tokenizer_path))
47
+ evidence = {'tokenizer': {'path': 'tokenizer/tokenizer.model', 'sha256': sha(tokenizer_path),
48
+ 'vocab_size': sp.vocab_size(),
49
+ 'ids': {'pad': sp.pad_id(), 'bos': sp.bos_id(), 'eos': sp.eos_id(), 'unk': sp.unk_id()}},
50
+ 'checks': []}
51
+
52
+ split_records = {s: read_jsonl(root / f'data/{s}.jsonl.gz') for s in SPLITS}
53
+ by_id = {r['id']: r for rows in split_records.values() for r in rows}
54
+ assigned = json.loads((root / 'audit/group-splits.json').read_text())['splits']
55
+
56
+ # 1. Split separation read straight from the emitted files, not from the audit report.
57
+ groups_from_files = {}
58
+ for split, rows in split_records.items():
59
+ for r in rows:
60
+ groups_from_files.setdefault(r['group'], set()).add(split)
61
+ overlapping = {g: sorted(s) for g, s in groups_from_files.items() if len(s) > 1}
62
+ absent = sorted(g for g in assigned if g not in groups_from_files)
63
+ evidence['checks'].append({
64
+ 'name': 'split_separation_from_emitted_files',
65
+ 'groups_in_files': len(groups_from_files),
66
+ 'groups_with_multiple_splits': overlapping,
67
+ 'groups_in_audit_but_absent_from_files': absent,
68
+ 'split_record_counts': {s: len(v) for s, v in split_records.items()},
69
+ 'split_group_counts': {s: len({r['group'] for r in v}) for s, v in split_records.items()},
70
+ 'passed': not overlapping and not absent,
71
+ })
72
+
73
+ # 2. Decode whole packed rows back through the SentencePiece model.
74
+ packed = root / 'packed'
75
+ manifest = json.loads((packed / 'manifest.json').read_text())
76
+ decode_cases = []
77
+ partial_rows = []
78
+ for split in SPLITS:
79
+ tokens_meta = read_jsonl(packed / f'{split}-records.jsonl')
80
+ starts = [0]
81
+ for t in tokens_meta:
82
+ starts.append(starts[-1] + t['tokens'])
83
+ candidates = []
84
+ for index, entry in enumerate(tokens_meta):
85
+ lo, hi = starts[index], starts[index + 1]
86
+ row = lo // 2048
87
+ if hi <= (row + 1) * 2048:
88
+ candidates.append((entry['id'], row, lo - row * 2048, hi - row * 2048))
89
+ candidates.sort(key=lambda c: (c[3] - c[2]))
90
+ # Sample both extremes among records that fit inside one row, and force one Lean case per split.
91
+ chosen = candidates[:2] + candidates[-2:]
92
+ lean = [(rid, row, lo, hi) for rid, row, lo, hi in candidates
93
+ if by_id[rid]['record_type'] == 'formal_lean_proof']
94
+ if lean:
95
+ chosen.append(max(lean, key=lambda c: c[3] - c[2]))
96
+ if split == 'dev':
97
+ chosen = candidates
98
+ if not chosen:
99
+ # No record fits inside one row. Compare the trailing partial row against the token stream of the
100
+ # record it covers, decoded from that record's own tokens.
101
+ tokens_meta_rows = read_jsonl(packed / f'{split}-rows.jsonl')
102
+ last = tokens_meta_rows[-1]
103
+ record_id = last['segments'][-1]['record_id']
104
+ record = by_id[record_id]
105
+ record_tokens = [1] + sp.encode(record['text'], out_type=int) + [2]
106
+ lo = last['segments'][0]['record_start']
107
+ hi = last['segments'][-1]['record_end']
108
+ expected_tokens = record_tokens[lo:hi]
109
+ shard_entry = manifest['splits'][split][last['shard']]
110
+ ids = np.load(packed / shard_entry['fields']['input_ids']['path'], mmap_mode='r', allow_pickle=False)
111
+ n = last['length']
112
+ array = np.asarray(ids[last['shard_row'], :n], dtype=np.int64).tolist()
113
+ decoded = sp.decode(array)
114
+ expected_text = sp.decode(expected_tokens)
115
+ partial_rows.append({
116
+ 'split': split, 'packed_row': last['row'], 'record_id': record_id,
117
+ 'record_type': record['record_type'], 'records_in_row': len(last['segments']),
118
+ 'tokens': n, 'record_token_slice': [lo, hi],
119
+ 'packed_tokens_equal_record_token_slice': array == expected_tokens,
120
+ 'decoded_text_equal_expected_slice_decode': decoded == expected_text,
121
+ 'decoded_characters': len(decoded), 'source_characters': len(record['text']),
122
+ 'whitespace_removed_text_match': re.sub(r'\s+', '', decoded) ==
123
+ re.sub(r'\s+', '', expected_text),
124
+ 'decoded_preview': decoded[:100].replace('\n', '\\n'),
125
+ })
126
+ seen_rows = set()
127
+ for record_id, row, lo, hi in chosen:
128
+ if (row, lo, hi) in seen_rows:
129
+ continue
130
+ seen_rows.add((row, lo, hi))
131
+ shard_index = row // 256
132
+ shard_row = row - shard_index * 256
133
+ shard = manifest['splits'][split][shard_index]
134
+ ids = np.load(packed / shard['fields']['input_ids']['path'], mmap_mode='r', allow_pickle=False)
135
+ array = np.asarray(ids[shard_row, lo:hi], dtype=np.int64).tolist()
136
+ decoded = sp.decode(array)
137
+ record = by_id[record_id]
138
+ recoded = [1] + sp.encode(decoded, out_type=int) + [2]
139
+ decode_cases.append({
140
+ 'split': split, 'packed_row': row, 'column_start': lo, 'column_end': hi,
141
+ 'record_id': record_id, 'record_type': record['record_type'],
142
+ 'tokens': int(hi - lo), 'bos_ok': array[0] == 1, 'eos_ok': array[-1] == 2,
143
+ 'decoded_reencodes_to_same_ids': recoded == array,
144
+ 'byte_exact_text_match': decoded == record['text'],
145
+ 'whitespace_normalized_text_match': normalize(decoded) == normalize(record['text']),
146
+ 'decoded_characters': len(decoded), 'source_characters': len(record['text']),
147
+ 'whitespace_removed_text_match': re.sub(r'\s+', '', decoded) == re.sub(r'\s+', '', record['text']),
148
+ 'decoded_preview': decoded[:100].replace('\n', '\\n'),
149
+ })
150
+ evidence['decode_cases'] = decode_cases
151
+ evidence['partial_row_cases'] = partial_rows
152
+ evidence['checks'].insert(1, {
153
+ 'name': 'packed_rows_decode_to_source_text',
154
+ 'cases': len(decode_cases),
155
+ 'byte_exact_all': all(c['byte_exact_text_match'] for c in decode_cases),
156
+ 'whitespace_normalized_all': all(c['whitespace_normalized_text_match'] for c in decode_cases),
157
+ 'non_whitespace_characters_identical_all': all(c['whitespace_removed_text_match'] for c in decode_cases),
158
+ 'reencode_identical_all': all(c['decoded_reencodes_to_same_ids'] for c in decode_cases),
159
+ 'bos_eos_all': all(c['bos_ok'] and c['eos_ok'] for c in decode_cases),
160
+ 'record_types_covered': sorted({c['record_type'] for c in decode_cases}),
161
+ 'passed': all(c['whitespace_removed_text_match'] and c['decoded_reencodes_to_same_ids']
162
+ and c['bos_ok'] and c['eos_ok'] for c in decode_cases),
163
+ })
164
+ evidence['checks'].insert(2, {
165
+ 'name': 'partial_rows_matching_record_token_slices',
166
+ 'cases': len(partial_rows),
167
+ 'passed': all(c['packed_tokens_equal_record_token_slice']
168
+ and c['decoded_text_equal_expected_slice_decode'] for c in partial_rows),
169
+ })
170
+
171
+ # 3. One packed row that covers several records, decoded and compared piecewise.
172
+ split = 'train'
173
+ rows_meta = read_jsonl(packed / f'{split}-rows.jsonl')
174
+ for entry in rows_meta:
175
+ if entry['segments'][0]['record_start'] == 0 and len(entry['segments']) > 1 \
176
+ and entry['segments'][-1]['record_end'] == entry['length'] - 1:
177
+ shard = manifest['splits'][split][entry['shard']]
178
+ ids = np.load(packed / shard['fields']['input_ids']['path'], mmap_mode='r', allow_pickle=False)
179
+ array = np.asarray(ids[entry['shard_row'], :entry['length']], dtype=np.int64).tolist()
180
+ decoded = sp.decode(array)
181
+ joined = normalize(''.join(by_id[s['record_id']]['text'][s['record_start']:s['record_end']]
182
+ for s in entry['segments']))
183
+ evidence['checks'].append({
184
+ 'name': 'packed_row_decode_covering_several_records',
185
+ 'split': split, 'packed_row': entry['row'], 'records': len(entry['segments']),
186
+ 'tokens': entry['length'], 'whitespace_normalized_text_match': normalize(decoded) == joined,
187
+ 'record_types': sorted({by_id[s['record_id']]['record_type'] for s in entry['segments']}),
188
+ 'decoded_preview': decoded[:100].replace('\n', '\\n'),
189
+ 'passed': normalize(decoded) == joined,
190
+ })
191
+ break
192
+
193
+ # 4. Parquet row round-trips to its JSONL record.
194
+ import pyarrow.parquet as pq
195
+ roundtrips = []
196
+ for split in SPLITS:
197
+ table = pq.read_table(root / f'data/{split}.parquet')
198
+ rows = split_records[split]
199
+ for row_index in (0, len(rows) // 2, len(rows) - 1):
200
+ row = table.slice(row_index, 1).to_pylist()[0]
201
+ metadata = json.loads(row.pop('metadata_json'))
202
+ rebuilt = {**row, **metadata}
203
+ original = rows[row_index]
204
+ roundtrips.append({
205
+ 'split': split, 'parquet_row': row_index, 'record_id': original['id'],
206
+ 'columns': table.column_names,
207
+ 'byte_identical_canonical_json': canonical(rebuilt) == canonical(original),
208
+ 'text_sha256_matches': hashlib.sha256(rebuilt['text'].encode()).hexdigest() == original['text_sha256'],
209
+ })
210
+ evidence['checks'].append({'name': 'parquet_to_jsonl_roundtrip', 'cases': roundtrips,
211
+ 'passed': all(c['byte_identical_canonical_json'] for c in roundtrips)})
212
+
213
+ # 5. Exact-text uniqueness across all delivered records.
214
+ hashes = {hashlib.sha256(r['text'].encode()).hexdigest() for r in by_id.values()}
215
+ evidence['checks'].append({'name': 'cross_split_text_uniqueness', 'records': len(by_id),
216
+ 'unique_text_hashes': len(hashes), 'passed': len(hashes) == len(by_id)})
217
+
218
+ all_passed = all(c['passed'] for c in evidence['checks'])
219
+ evidence['status'] = 'passed' if all_passed else 'failed'
220
+ evidence['measured'] = {'records': len(by_id), 'groups': len(assigned),
221
+ 'splits': {s: len(v) for s, v in split_records.items()},
222
+ 'packed': manifest['statistics']}
223
+ path = args.write if args.write.is_absolute() else root / args.write
224
+ path.parent.mkdir(parents=True, exist_ok=True)
225
+ path.write_text(json.dumps(evidence, ensure_ascii=False, indent=2) + '\n', encoding='utf-8')
226
+ print(json.dumps(evidence, ensure_ascii=False, indent=2))
227
+ print(f'\nEvidence written to {path}')
228
+ if not all_passed:
229
+ raise SystemExit(1)
230
+
231
+
232
+ if __name__ == '__main__':
233
+ main()
components/google-alphaproof-nexus/tools/validate_google_alphaproof_nexus.py ADDED
@@ -0,0 +1,281 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ #!/usr/bin/env python3
2
+ """Validate the Google AlphaProof Nexus contribution transport, provenance, leakage controls and packed tokens."""
3
+ from __future__ import annotations
4
+ import argparse
5
+ import gzip
6
+ import hashlib
7
+ import json
8
+ import re
9
+ from pathlib import Path
10
+
11
+ CORE = ('id', 'text', 'group', 'license', 'paper_id', 'split', 'record_type', 'source_url')
12
+ COMMIT = '0647711a71183c1ea492ad60860776617ce1ea88'
13
+ SPLITS = ('train', 'dev', 'test')
14
+ RECORD_TYPES = {'formal_lean_proof', 'natural_language_proof'}
15
+ COPYRIGHT = re.compile(r'Copyright \d{4} Google LLC')
16
+
17
+
18
+ def ensure(condition, message):
19
+ if not condition:
20
+ raise ValueError(message)
21
+
22
+
23
+ def sha(path):
24
+ h = hashlib.sha256()
25
+ with Path(path).open('rb') as f:
26
+ for chunk in iter(lambda: f.read(1024 * 1024), b''):
27
+ h.update(chunk)
28
+ return h.hexdigest()
29
+
30
+
31
+ def read_jsonl(path):
32
+ path = Path(path)
33
+ opener = gzip.open if path.suffix == '.gz' else open
34
+ with opener(path, 'rt', encoding='utf-8') as f:
35
+ return [json.loads(line) for line in f if line.strip()]
36
+
37
+
38
+ def read_json(path):
39
+ return json.loads(Path(path).read_text(encoding='utf-8'))
40
+
41
+
42
+ def safe_path(root, relative):
43
+ ensure(not Path(relative).is_absolute(), 'Absolute manifest path')
44
+ p = (root / relative).resolve()
45
+ ensure(p.is_relative_to(root.resolve()), 'Path escapes component root')
46
+ return p
47
+
48
+
49
+ def check_parquet(path, expected):
50
+ import pyarrow.parquet as pq
51
+ table = pq.read_table(path)
52
+ ensure(table.column_names == [*CORE, 'metadata_json'], f'Parquet columns: {path}')
53
+ ensure(table.num_rows == len(expected), f'Parquet row count: {path}')
54
+ for index, (row, original) in enumerate(zip(table.to_pylist(), expected)):
55
+ metadata = json.loads(row.pop('metadata_json'))
56
+ ensure({**row, **metadata} == original, f'Parquet mismatch {path}:{index}')
57
+
58
+
59
+ def validate_packed(root, split_records):
60
+ import numpy as np
61
+ import sentencepiece as spm
62
+ packed = root / 'packed'
63
+ manifest = read_json(packed / 'manifest.json')
64
+ ensure(manifest['format'] == 'mmla-type2-packed-v1', 'Packed format')
65
+ ensure(manifest['seq_len'] == 2048, 'Packed sequence length')
66
+ ensure(manifest['source'] == 'google-alphaproof-nexus', 'Packed source')
67
+ tokenizer = (packed / manifest['tokenizer']['path']).resolve()
68
+ ensure(tokenizer.is_relative_to(root), 'Tokenizer path escapes component')
69
+ ensure(sha(tokenizer) == manifest['tokenizer']['sha256'], 'Tokenizer hash')
70
+ sp = spm.SentencePieceProcessor(model_file=str(tokenizer))
71
+ ensure((sp.pad_id(), sp.bos_id(), sp.eos_id(), sp.unk_id(), sp.vocab_size()) == (0, 1, 2, 3, 32768), 'Tokenizer IDs')
72
+ result = {}
73
+ for split in SPLITS:
74
+ records = split_records[split]
75
+ token_catalog = read_jsonl(packed / f'{split}-records.jsonl')
76
+ rows = read_jsonl(packed / f'{split}-rows.jsonl')
77
+ ensure([r['id'] for r in records] == [r['id'] for r in token_catalog], f'{split} packed source ordering')
78
+ arrays, doc_offsets, stream_start = [], [], 0
79
+ for r, entry in zip(records, token_catalog):
80
+ tokens = np.asarray([1] + sp.encode(r['text'], out_type=int) + [2], dtype='<i4')
81
+ ensure(len(tokens) == entry['tokens'], f'{split} record token count')
82
+ ensure(hashlib.sha256(tokens.tobytes()).hexdigest() == entry['token_sha256'], f'{split} record token hash')
83
+ ensure(r['paper_id'] == entry['paper_id'], f'{split} record paper ID')
84
+ doc_offsets.append((stream_start, stream_start + len(tokens), r['id'], r['paper_id']))
85
+ stream_start += len(tokens)
86
+ arrays.append(tokens)
87
+ stream = np.concatenate(arrays)
88
+ del arrays
89
+ cursor, doc_index, row_number, labels_count = 0, 0, 0, 0
90
+ for shard_index, shard in enumerate(manifest['splits'][split]):
91
+ loaded = {}
92
+ for name in ('input_ids', 'labels', 'role_ids'):
93
+ field = shard['fields'][name]
94
+ path = safe_path(packed, field['path'])
95
+ ensure(path.stat().st_size == field['bytes'] and sha(path) == field['sha256'], f'Packed file hash/size: {path}')
96
+ array = np.load(path, mmap_mode='r', allow_pickle=False)
97
+ ensure(array.dtype == np.dtype('int32'), 'Packed dtype')
98
+ ensure(list(array.shape) == field['shape'] == [shard['count'], 2048], 'Packed shape')
99
+ loaded[name] = array
100
+ ensure(bool(np.all(loaded['role_ids'] == 0)), 'Nonzero plain-text role')
101
+ ensure(int(loaded['input_ids'].min()) >= 0, 'Negative input token')
102
+ for shard_row in range(shard['count']):
103
+ ensure(row_number < len(rows), 'Missing row attribution')
104
+ attribution = rows[row_number]
105
+ n = min(2048, len(stream) - cursor)
106
+ ensure(n > 0, 'Unexpected all-padding row')
107
+ expect_input = np.zeros(2048, dtype=np.int32)
108
+ expect_input[:n] = stream[cursor:cursor + n]
109
+ expect_labels = np.full(2048, -100, dtype=np.int32)
110
+ available = min(n, len(stream) - cursor - 1)
111
+ if available > 0:
112
+ expect_labels[:available] = stream[cursor + 1:cursor + 1 + available]
113
+ expect_labels[expect_labels == 1] = -100
114
+ ensure(np.array_equal(loaded['input_ids'][shard_row], expect_input), f'{split} source token mismatch at row {row_number}')
115
+ ensure(np.array_equal(loaded['labels'][shard_row], expect_labels), f'{split} label shift/mask mismatch at row {row_number}')
116
+ ensure(attribution['row'] == row_number and attribution['shard'] == shard_index and attribution['shard_row'] == shard_row, 'Row location attribution')
117
+ ensure(attribution['stream_start'] == cursor and attribution['length'] == n, 'Row stream attribution')
118
+ ensure(attribution['segments'], 'Row without a source segment')
119
+ while doc_index < len(doc_offsets) and doc_offsets[doc_index][1] <= cursor:
120
+ doc_index += 1
121
+ spans, k = [], doc_index
122
+ while k < len(doc_offsets) and doc_offsets[k][0] < cursor + n:
123
+ start, end, rid, pid = doc_offsets[k]
124
+ lo, hi = max(cursor, start), min(cursor + n, end)
125
+ if lo < hi:
126
+ spans.append({'record_id': rid, 'paper_id': pid,
127
+ 'row_start': lo - cursor, 'row_end': hi - cursor,
128
+ 'record_start': lo - start, 'record_end': hi - start})
129
+ k += 1
130
+ ensure(attribution['segments'] == spans, 'Token source attribution mismatch')
131
+ labels_count += int(np.count_nonzero(expect_labels != -100))
132
+ cursor += n
133
+ row_number += 1
134
+ ensure(cursor == len(stream), f'{split} token stream not fully consumed')
135
+ ensure(row_number == len(rows), f'{split} extra row attribution')
136
+ actual = {'rows': row_number, 'tokens': len(stream), 'loss_bearing_labels': labels_count}
137
+ ensure(actual == manifest['statistics'][split], f'{split} packed statistics mismatch')
138
+ result[split] = actual
139
+ print(f'Validated packed {split}: {row_number} rows, {len(stream)} tokens', flush=True)
140
+ return result
141
+
142
+
143
+ def main():
144
+ parser = argparse.ArgumentParser(description=__doc__)
145
+ parser.add_argument('--component', type=Path, default=Path(__file__).resolve().parents[1])
146
+ parser.add_argument('--source-root', type=Path, help='Pinned upstream checkout to hash-check source files')
147
+ args = parser.parse_args()
148
+ root = args.component.resolve()
149
+ checks = []
150
+ manifest = read_json(root / 'manifest.json')
151
+ ensure(manifest['format'] == 'cayley-contribution-file-manifest-v1', 'Manifest format')
152
+ ignored = set(manifest['excludes'])
153
+ expected_files = {x['path'] for x in manifest['files']}
154
+ actual_files = {p.relative_to(root).as_posix() for p in root.rglob('*')
155
+ if p.is_file() and p.relative_to(root).as_posix() not in ignored}
156
+ ensure(expected_files == actual_files,
157
+ f'File manifest coverage mismatch: missing={sorted(expected_files - actual_files)}, extra={sorted(actual_files - expected_files)}')
158
+ ensure(not any('__pycache__' in p or p.endswith('.pyc') for p in actual_files), 'Build artefacts in component')
159
+ ensure(not any('.DS_Store' in p for p in actual_files), '.DS_Store in component')
160
+ for item in manifest['files']:
161
+ path = safe_path(root, item['path'])
162
+ ensure(path.stat().st_size == item['bytes'], f'Byte count: {path}')
163
+ ensure(sha(path) == item['sha256'], f'File digest: {path}')
164
+ checks.append('all_delivered_file_hashes_and_sizes')
165
+
166
+ sources = read_json(root / 'sources.json')
167
+ upstream = sources['upstream']
168
+ ensure(upstream['commit'] == COMMIT, 'sources.json commit')
169
+ ensure(upstream['license'] == 'Apache-2.0' and sources['all_records_license'] == 'Apache-2.0', 'sources.json license')
170
+ ensure(upstream['creators'] == ['Google DeepMind'], 'sources.json creators')
171
+ ensure(sources['mathematical_verification'] == 'not_performed', 'sources.json verification note')
172
+ ensure(upstream['fixed_revision_url'] == f'https://github.com/google-deepmind/alphaproof-nexus-results/tree/{COMMIT}', 'fixed revision url')
173
+ checks.append('pinned_sources_metadata')
174
+
175
+ source_files = read_jsonl(root / 'audit/source-files.jsonl')
176
+ ensure(source_files, 'Empty source file inventory')
177
+ ensure(len({x['path'] for x in source_files}) == len(source_files), 'Duplicate source inventory path')
178
+ ensure(all(x['source_url'] == f'https://github.com/google-deepmind/alphaproof-nexus-results/blob/{COMMIT}/{x["path"]}'
179
+ for x in source_files), 'Source inventory URL not pinned')
180
+ if args.source_root:
181
+ for item in source_files:
182
+ path = safe_path(args.source_root.resolve(), item['path'])
183
+ ensure(path.is_file(), f'Missing upstream file: {item["path"]}')
184
+ ensure(path.stat().st_size == item['bytes'], f'Upstream size: {item["path"]}')
185
+ ensure(sha(path) == item['sha256'], f'Upstream digest: {item["path"]}')
186
+ checks.append('source_file_inventory' + ('_local_bytes' if args.source_root else '_metadata_only'))
187
+
188
+ group_splits = read_json(root / 'audit/group-splits.json')
189
+ assigned = group_splits['splits']
190
+ ensure(set(assigned.values()) == set(SPLITS), f'Splits present: {set(assigned.values())}')
191
+ ensure(len(assigned) >= 3, 'Too few groups for leakage-safe splits')
192
+ split_records, ids, hashes, groups_seen = {}, set(), set(), {}
193
+ source_by_path = {x['path']: x for x in source_files}
194
+ for split in SPLITS:
195
+ rows = read_jsonl(root / f'data/{split}.jsonl.gz')
196
+ ensure(rows, f'Empty split {split}')
197
+ check_parquet(root / f'data/{split}.parquet', rows)
198
+ for r in rows:
199
+ ensure(r['split'] == split and assigned[r['group']] == split, 'Group split assignment')
200
+ ensure(r['record_type'] in RECORD_TYPES, f'Record type: {r["id"]}')
201
+ ensure(r['license'] == 'Apache-2.0', 'Record license')
202
+ ensure(r['source_commit'] == COMMIT and COMMIT in r['source_url'], 'Record pinned commit')
203
+ ensure(r['source_url'].endswith(r['source_path']), 'Record source URL/path mismatch')
204
+ ensure(r['content_origin'] in {'source_verbatim', 'source_pdf_extraction'}, 'Content origin')
205
+ ensure(r['mathematical_verification'] == 'not_performed' and r['lean_verified'] is False, 'Verification claim')
206
+ ensure(r['creators'] == ['Google DeepMind'], 'Record creators')
207
+ ensure(r['paper_id'] == r['group'], 'paper_id is the problem group')
208
+ ensure(r['id'] not in ids, 'Cross-split duplicate ID')
209
+ ids.add(r['id'])
210
+ digest = hashlib.sha256(r['text'].encode('utf-8')).hexdigest()
211
+ ensure(r['text_sha256'] == digest and digest not in hashes, 'Exact text duplicate or digest mismatch')
212
+ hashes.add(digest)
213
+ ensure(r['source_path'] in source_by_path, 'Record source path absent from inventory')
214
+ ensure(source_by_path[r['source_path']]['sha256'] == r['source_sha256'], 'Record source digest mismatch')
215
+ if r['record_type'] == 'formal_lean_proof':
216
+ ensure(r['extraction_quality'] == 'verbatim_source_text', 'Lean extraction quality')
217
+ ensure(r['transform'] == 'identity_verbatim_bytes', 'Lean transform')
218
+ ensure(bool(COPYRIGHT.search(r['text'][:600])), f'Retained copyright header: {r["source_path"]}')
219
+ if args.source_root:
220
+ original = safe_path(args.source_root.resolve(), r['source_path']).read_text(encoding='utf-8')
221
+ ensure(r['text'] == original, f'Lean text differs from pinned bytes: {r["source_path"]}')
222
+ else:
223
+ ensure(r['extraction_quality'] == 'lossy_pdf_text_layer', 'PDF extraction quality flag')
224
+ ensure(r['extraction_metrics']['pages'] >= 1, 'PDF page count')
225
+ ensure(r['extraction_metrics']['characters'] == len(r['text']), 'PDF character count')
226
+ if r['group'] in groups_seen:
227
+ ensure(groups_seen[r['group']] == split, 'Group cross-split leakage')
228
+ groups_seen[r['group']] = split
229
+ split_records[split] = rows
230
+ checks.append('lossless_jsonl_parquet_and_exact_text_uniqueness')
231
+ checks.append('problem_group_split_separation')
232
+ checks.append('verbatim_lean_source_equality' + ('_local_bytes' if args.source_root else '_metadata_only'))
233
+
234
+ pairing = {}
235
+ for r in sum(split_records.values(), []):
236
+ pairing.setdefault(r['group'], set()).add(r['record_type'])
237
+ paired = sorted(g for g, t in pairing.items() if t == RECORD_TYPES)
238
+ ensure(len(paired) >= 10, 'Too few paired groups')
239
+ checks.append('lean_and_prose_companions_share_group')
240
+
241
+ extraction = read_jsonl(root / 'audit/extraction.jsonl')
242
+ prose = {r['source_path']: r for r in sum(split_records.values(), []) if r['record_type'] == 'natural_language_proof'}
243
+ ensure({x['source_path'] for x in extraction} == set(prose), 'Extraction audit coverage')
244
+ for entry in extraction:
245
+ ensure(entry['extraction_quality'] == 'lossy_pdf_text_layer', 'Extraction audit quality flag')
246
+ ensure(entry['pages'] >= 1 and len(entry['page_characters']) == entry['pages'], 'Per-page counts')
247
+ ensure(entry['characters'] == prose[entry['source_path']]['extraction_metrics']['characters'], 'Extraction character count')
248
+ ensure(all(v >= 0 for v in entry['artifact_characters'].values()), 'Negative artifact count')
249
+ checks.append('per_pdf_extraction_audit')
250
+
251
+ duplicates = read_jsonl(root / 'audit/excluded-duplicates.jsonl')
252
+ ensure(len({d['excluded_id'] for d in duplicates}) == len(duplicates), 'Duplicate excluded IDs')
253
+ checks.append('dedup_audit_present')
254
+
255
+ packed = validate_packed(root, split_records) if (root / 'packed/manifest.json').exists() else {}
256
+ if packed:
257
+ checks.append('all_packed_tokens_reencoded_source_shift_padding_masks_and_attribution')
258
+ stats = read_json(root / 'statistics.json')
259
+ ensure(stats['records'] == sum(len(v) for v in split_records.values()), 'Statistics record count')
260
+ ensure(stats['records'] == stats['records_before_deduplication'] - stats['excluded_exact_duplicates'], 'Statistics dedup accounting')
261
+ ensure(stats['packed'] == packed, 'Statistics packed counts')
262
+ ensure(stats['problem_groups'] == len(assigned), 'Statistics group count')
263
+ for split in SPLITS:
264
+ ensure(stats['splits'][split]['records'] == len(split_records[split]), f'Statistics {split} records')
265
+ checks.append('statistics_match_artifacts')
266
+ report = {'status': 'passed', 'source_commit': COMMIT, 'records': stats['records'],
267
+ 'problem_groups': stats['problem_groups'], 'record_types': stats['record_types'],
268
+ 'splits': stats['splits'], 'paired_groups': len(paired),
269
+ 'source_files_hashed': len(source_files), 'packed': packed, 'checks': checks,
270
+ 'mathematical_verification': 'not_performed',
271
+ 'limitations': ['PDF extraction is lossy for symbols, ligatures and formula layout.',
272
+ 'Upstream contains only problems solved by AlphaProof Nexus; the collection is selection-biased.',
273
+ 'No Lean compilation, proof checking or mathematical verification.',
274
+ 'No near-duplicate or Phase 1 overlap assessment.',
275
+ 'SentencePiece decoding normalizes runs of spaces, inherited from the shared Cayley-Math tokenizer.']}
276
+ (root / 'validation.json').write_text(json.dumps(report, ensure_ascii=False, indent=2) + '\n', encoding='utf-8')
277
+ print(json.dumps(report, ensure_ascii=False, indent=2))
278
+
279
+
280
+ if __name__ == '__main__':
281
+ main()
components/google-alphaproof-nexus/validation.json ADDED
@@ -0,0 +1,78 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "status": "passed",
3
+ "source_commit": "0647711a71183c1ea492ad60860776617ce1ea88",
4
+ "records": 83,
5
+ "problem_groups": 51,
6
+ "record_types": {
7
+ "formal_lean_proof": 71,
8
+ "natural_language_proof": 12
9
+ },
10
+ "splits": {
11
+ "train": {
12
+ "records": 80,
13
+ "problems": 49,
14
+ "text_characters": 2154974,
15
+ "record_types": {
16
+ "formal_lean_proof": 69,
17
+ "natural_language_proof": 11
18
+ }
19
+ },
20
+ "dev": {
21
+ "records": 1,
22
+ "problems": 1,
23
+ "text_characters": 6805,
24
+ "record_types": {
25
+ "formal_lean_proof": 1
26
+ }
27
+ },
28
+ "test": {
29
+ "records": 2,
30
+ "problems": 1,
31
+ "text_characters": 67992,
32
+ "record_types": {
33
+ "formal_lean_proof": 1,
34
+ "natural_language_proof": 1
35
+ }
36
+ }
37
+ },
38
+ "paired_groups": 10,
39
+ "source_files_hashed": 92,
40
+ "packed": {
41
+ "train": {
42
+ "rows": 491,
43
+ "tokens": 1004469,
44
+ "loss_bearing_labels": 1004389
45
+ },
46
+ "dev": {
47
+ "rows": 2,
48
+ "tokens": 3144,
49
+ "loss_bearing_labels": 3143
50
+ },
51
+ "test": {
52
+ "rows": 14,
53
+ "tokens": 28411,
54
+ "loss_bearing_labels": 28409
55
+ }
56
+ },
57
+ "checks": [
58
+ "all_delivered_file_hashes_and_sizes",
59
+ "pinned_sources_metadata",
60
+ "source_file_inventory_local_bytes",
61
+ "lossless_jsonl_parquet_and_exact_text_uniqueness",
62
+ "problem_group_split_separation",
63
+ "verbatim_lean_source_equality_local_bytes",
64
+ "lean_and_prose_companions_share_group",
65
+ "per_pdf_extraction_audit",
66
+ "dedup_audit_present",
67
+ "all_packed_tokens_reencoded_source_shift_padding_masks_and_attribution",
68
+ "statistics_match_artifacts"
69
+ ],
70
+ "mathematical_verification": "not_performed",
71
+ "limitations": [
72
+ "PDF extraction is lossy for symbols, ligatures and formula layout.",
73
+ "Upstream contains only problems solved by AlphaProof Nexus; the collection is selection-biased.",
74
+ "No Lean compilation, proof checking or mathematical verification.",
75
+ "No near-duplicate or Phase 1 overlap assessment.",
76
+ "SentencePiece decoding normalizes runs of spaces, inherited from the shared Cayley-Math tokenizer."
77
+ ]
78
+ }