Upload 21 Lean 4 datasets with v4.21.0 verification

README.md

@@ -0,0 +1,240 @@
+---
+license: apache-2.0
+task_categories:
+  - text-generation
+language:
+  - en
+tags:
+  - lean4
+  - theorem-proving
+  - formal-mathematics
+  - verification
+  - math
+size_categories:
+  - 1M<n<10M
+configs:
+  - config_name: compfiles
+    data_files:
+      - split: train
+        path: data/compfiles/train-*.parquet
+  - config_name: deepseek_prover
+    data_files:
+      - split: train
+        path: data/deepseek_prover/train-*.parquet
+  - config_name: deepseek_proverbench
+    data_files:
+      - split: train
+        path: data/deepseek_proverbench/train-*.parquet
+  - config_name: formal_conjectures
+    data_files:
+      - split: train
+        path: data/formal_conjectures/train-*.parquet
+  - config_name: formalmath
+    data_files:
+      - split: train
+        path: data/formalmath/train-*.parquet
+  - config_name: goedel_minif2f
+    data_files:
+      - split: train
+        path: data/goedel_minif2f/train-*.parquet
+  - config_name: goedel_mobench
+    data_files:
+      - split: train
+        path: data/goedel_mobench/train-*.parquet
+  - config_name: goedel_pset
+    data_files:
+      - split: train
+        path: data/goedel_pset/train-*.parquet
+  - config_name: goedel_test
+    data_files:
+      - split: train
+        path: data/goedel_test/train-*.parquet
+  - config_name: hf_lean_workbook
+    data_files:
+      - split: train
+        path: data/hf_lean_workbook/train-*.parquet
+  - config_name: hf_minif2f_lean4
+    data_files:
+      - split: train
+        path: data/hf_minif2f_lean4/train-*.parquet
+  - config_name: hf_minif2f_v2
+    data_files:
+      - split: train
+        path: data/hf_minif2f_v2/train-*.parquet
+  - config_name: hf_tonic_minif2f
+    data_files:
+      - split: train
+        path: data/hf_tonic_minif2f/train-*.parquet
+  - config_name: lean_proofs
+    data_files:
+      - split: train
+        path: data/lean_proofs/train-*.parquet
+  - config_name: lean_workbook_full
+    data_files:
+      - split: train
+        path: data/lean_workbook_full/train-*.parquet
+  - config_name: matholympiadbench
+    data_files:
+      - split: train
+        path: data/matholympiadbench/train-*.parquet
+  - config_name: nemotron_proofs
+    data_files:
+      - split: train
+        path: data/nemotron_proofs/train-*.parquet
+  - config_name: numinamath_lean
+    data_files:
+      - split: train
+        path: data/numinamath_lean/train-*.parquet
+  - config_name: proofnet
+    data_files:
+      - split: train
+        path: data/proofnet/train-*.parquet
+  - config_name: putnam2025
+    data_files:
+      - split: train
+        path: data/putnam2025/train-*.parquet
+  - config_name: putnam_bench
+    data_files:
+      - split: train
+        path: data/putnam_bench/train-*.parquet
+dataset_info:
+  features:
+    - name: id
+      dtype: string
+    - name: source
+      dtype: string
+    - name: formal_statement
+      dtype: string
+    - name: header
+      dtype: string
+    - name: lean4_code
+      dtype: string
+    - name: has_proof
+      dtype: bool
+    - name: proof_body
+      dtype: string
+    - name: natural_language
+      dtype: string
+    - name: lean_version
+      dtype: string
+    - name: split
+      dtype: string
+    - name: tags
+      sequence: string
+    - name: category
+      dtype: string
+    - name: verification
+      dtype: string
+    - name: v4210_is_valid
+      dtype: bool
+    - name: v4210_compiles
+      dtype: bool
+    - name: v4210_has_sorry
+      dtype: bool
+    - name: v4210_latency_s
+      dtype: float64
+---
+
+# ai4math-lean
+
+21 Lean 4 formal mathematics datasets with machine-verified labels.
+
+Every problem has been verified against lean-server v4.21.0 on UVA's HPC cluster.
+Results are embedded in each row as both structured `verification` JSON and flattened convenience columns.
+
+## Quick Start
+
+```python
+from datasets import load_dataset
+
+# Load a single dataset
+ds = load_dataset("charliemeyer2000/ai4math-lean", "deepseek_prover")
+
+# Load a large dataset with streaming
+ds = load_dataset("charliemeyer2000/ai4math-lean", "goedel_pset", streaming=True)
+
+# Filter to valid proofs
+valid = ds.filter(lambda x: x["v4210_is_valid"] == True)
+
+# Filter to RL targets (compiles but no proof)
+rl = ds.filter(lambda x: x["v4210_compiles"] and not x["has_proof"])
+```
+
+## Totals
+
+| Metric | Count |
+|--------|-------|
+| Total problems | 3,394,310 |
+| With proofs (SFT) | 1,274,168 |
+| Valid on v4.21.0 | 56,099 |
+
+## Datasets
+
+| Dataset | Problems | Proofs | Valid v4.21.0 |
+|---------|----------|--------|---------------|
+| compfiles | 297 | 259 | 44 |
+| deepseek_prover | 27,503 | 27,503 | 27,052 |
+| deepseek_proverbench | 325 | 325 | 0 |
+| formal_conjectures | 618 | 13 | 0 |
+| formalmath | 5,560 | 5,556 | 0 |
+| goedel_minif2f | 244 | 0 | 0 |
+| goedel_mobench | 360 | 0 | 0 |
+| goedel_pset | 1,732,594 | 263,589 | 3,193 |
+| goedel_test | 8 | 0 | 0 |
+| hf_lean_workbook | 29,750 | 29,750 | 25,448 |
+| hf_minif2f_lean4 | 231 | 0 | 0 |
+| hf_minif2f_v2 | 488 | 0 | 0 |
+| hf_tonic_minif2f | 488 | 0 | 0 |
+| lean_proofs | 98 | 97 | 0 |
+| lean_workbook_full | 140,214 | 0 | 0 |
+| matholympiadbench | 360 | 0 | 0 |
+| nemotron_proofs | 1,349,950 | 915,078 | 358 |
+| numinamath_lean | 104,155 | 31,615 | 0 |
+| proofnet | 371 | 371 | 0 |
+| putnam2025 | 24 | 12 | 4 |
+| putnam_bench | 672 | 0 | 0 |
+
+## Schema
+
+Each row contains:
+
+| Field | Type | Description |
+|-------|------|-------------|
+| `id` | string | Unique problem identifier |
+| `source` | string | Dataset source name |
+| `formal_statement` | string | Lean 4 theorem statement (no proof body) |
+| `header` | string | Import/setup code |
+| `lean4_code` | string | Full runnable Lean 4 code |
+| `has_proof` | bool | True if contains a real proof (not sorry) |
+| `proof_body` | string? | Proof tactic block (if has_proof) |
+| `natural_language` | string? | NL problem statement (if available) |
+| `lean_version` | string? | Original Lean toolchain version |
+| `split` | string? | train/test/val split |
+| `tags` | list[string] | Tags (topic, difficulty, etc.) |
+| `category` | string? | Math category |
+| `verification` | string (JSON) | Full verification results per Lean version |
+| `v4210_is_valid` | bool? | True=valid, False=errors/sorry, None=timeout |
+| `v4210_compiles` | bool | Whether the code compiled (is_valid is not None) |
+| `v4210_has_sorry` | bool? | Whether sorry was detected |
+| `v4210_latency_s` | float? | Verification wall time in seconds |
+
+## Verification Details
+
+The `verification` column contains a JSON string mapping Lean versions to results:
+
+```json
+{
+  "v4.21.0": {
+    "is_valid": false,
+    "has_sorry": true,
+    "errors": [],
+    "timeout_s": 600.0,
+    "latency_s": 34.59,
+    "verified_at": "2026-03-26T21:57:47Z"
+  }
+}
+```
+
+## License
+
+Apache 2.0

data/compfiles/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:457d056e9fcec0a2cffff93e1c6238c0151e10fbd4c0db470d2b134767484fbe
+size 3573471

data/deepseek_prover/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c2fb5e1c27b8fc0575cd16a4c2d223d57e4a071fe3bba3fb3a227201b8f81175
+size 8117307

data/deepseek_proverbench/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6e0a7dd612629d9dbd6f98c3815797a75eff606ecb1660ee7c170643e7197ef
+size 102850

data/formal_conjectures/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28df2cea42cf6d26a2770135d7bfe640fcd1952e2b8e4e96ebcd7bdb93924fa4
+size 14099213

data/formalmath/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1e1b7c8da94b5b53e5c31c0b87ff8f172e5c7aa28c88038855b2bcced1e2551
+size 2824865

data/goedel_minif2f/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5901ca8a260932b3db1e92a9362ceb7ebb95354e391c69efdf976072a454baa8
+size 150440

data/goedel_mobench/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6cbe6936bb0d52a70039d99e16ae060616556815d56ba837227417a4bb9ce3ee
+size 251743

data/goedel_pset/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58658b50a897499bf451ae847aea66c9d19d9329856e801745744375dea60718
+size 451903861

data/goedel_pset/train-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1c9a5d48099caa3d9e1c0689774f189bf327a16e82e8ae17531f250431f66e8
+size 451175707

data/goedel_pset/train-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9cbe9d88b65c83cb81c580f1e16037f636191d348b1cadcefd21153b34369319
+size 452130211

data/goedel_pset/train-00003.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f72044a2a87317e59a21f738e20d6527c72d2511cd57c1ea9541c7bde2ead7a5
+size 210616636

data/goedel_test/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f11b00aa0e8bdaff9d546fec853998a037e0f7c80e71e4e51b00236453247bf0
+size 16361

data/hf_lean_workbook/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:50dfc90ed492a52abb205a13b869442e67ac3564b006f04be0771b32983744e5
+size 46199289

data/hf_minif2f_lean4/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa59411994732c8875381b283469cceb9a78cdfdf2674a49a30906e3fbc577a4
+size 48016

data/hf_minif2f_v2/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:80aeda84ed8aa2bc041b8b5763af8a447c945ac1c202b917f4239c2449f3577e
+size 155904

data/hf_tonic_minif2f/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ca00f6402d1c8b1dda3a18d4e677dde85b26c56190fb00202c9aa8471ae87e9
+size 172920

data/lean_proofs/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3b4e56435fd5e45e3fc9a3040d4555c97492695016249ed67acb8b8d86158b87
+size 9119923

data/lean_workbook_full/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:971a6490987dbbece195d3b056860b0e9c826d5bf06062b64e7cbac0fcc03f0b
+size 37286623

data/matholympiadbench/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a0cf5ef2b13c7a628bd7914c67d27fe77ffc424ce341c0a1996f29b3948ea3cf
+size 152865

data/nemotron_proofs/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d513663f0d9be3d8e1677e6d7a45440f4bee9d5889c6520ef30d9541d058919e
+size 957045584

data/nemotron_proofs/train-00001.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3592e74093edfaf74e2d6a58676cd0cb35e94d20852b9d181f4c21b014306509
+size 961747934

data/nemotron_proofs/train-00002.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:53868034bd44fc1ad13b3f3f94e6e216b02e06a0a714a569b68e83f60b2ffa73
+size 680924828

data/numinamath_lean/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8aa277f7e67b7d41aa3e35ce761f4cbaa9df6ac3308f90960398e5144a5c19a2
+size 113791208

data/proofnet/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dfcae3ed6eef76aa149343a7fd26198bee4c9ace502b7215954d428af9d87a98
+size 104984

data/putnam2025/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25c21b62ceda97a0d2dcf4eaec563a53320b6cc987da8f6d255d9f7b678e2db8
+size 543677

data/putnam_bench/train-00000.parquet

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:005394ee6d72c88848fd773500a2e7f4d024de8eba1f969688b0619ae93452ff
+size 492555