Datasets:

alpha-proof-open-source
/

10-4-alpha-proof-trajectories

Error code:   StreamingRowsError
Exception:    CastError
Message:      Couldn't cast
schema: string
wave_id: string
created_at: timestamp[s]
store: string
shard_size: int64
counts: struct<searches: int64, searches_with_transitions: int64, searches_skipped_unsolved: int64, transiti (... 94 chars omitted)
  child 0, searches: int64
  child 1, searches_with_transitions: int64
  child 2, searches_skipped_unsolved: int64
  child 3, transitions_raw: int64
  child 4, dropped_focus: int64
  child 5, dropped_dedup: int64
  child 6, transitions: int64
  child 7, shards: int64
dedup: struct<enabled: bool, key: list<item: string>, dropped: int64>
  child 0, enabled: bool
  child 1, key: list<item: string>
      child 0, item: string
  child 2, dropped: int64
merge_logprobs: null
policy_version: string
policy_versions: list<item: string>
  child 0, item: string
versions: struct<lean: string, mathlib: string, reap: string, patchset_hash: string>
  child 0, lean: string
  child 1, mathlib: string
  child 2, reap: string
  child 3, patchset_hash: string
shards: list<item: struct<file: string, count: int64, bytes: int64, sha256: string>>
  child 0, item: struct<file: string, count: int64, bytes: int64, sha256: string>
      child 0, file: string
      child 1, count: int64
      child 2, bytes: int64
      child 3, sha256: string
sources: list<item: struct<search: string, problem_id: string, policy_version: string, transitions: int64>>
  child 0, item: struct<search: string, problem_id: string, policy_version: string, transitions: int64>
      child 0, search: string
      child 1, problem_id: string
      child 2, policy_version: string
      child 3, transitions: int64
sha256: string
dataset: string
schema_version: string
shard_format: string
updated_at: timestamp[s]
batches: list<item: struct<batch_id: string, path: string, manifest: string, shards: list<item: string>, prob (... 187 chars omitted)
  child 0, item: struct<batch_id: string, path: string, manifest: string, shards: list<item: string>, problems: int64 (... 175 chars omitted)
      child 0, batch_id: string
      child 1, path: string
      child 2, manifest: string
      child 3, shards: list<item: string>
          child 0, item: string
      child 4, problems: int64
      child 5, transitions: int64
      child 6, solved: int64
      child 7, terminal_verified: int64
      child 8, created_at: timestamp[s]
      child 9, versions: struct<lean: string, mathlib: string, reap: string, policy_version: string>
          child 0, lean: string
          child 1, mathlib: string
          child 2, reap: string
          child 3, policy_version: string
license: string
to
{'schema_version': Value('string'), 'dataset': Value('string'), 'updated_at': Value('timestamp[s]'), 'shard_format': Value('string'), 'license': Value('string'), 'batches': List({'batch_id': Value('string'), 'path': Value('string'), 'manifest': Value('string'), 'shards': List(Value('string')), 'problems': Value('int64'), 'transitions': Value('int64'), 'solved': Value('int64'), 'terminal_verified': Value('int64'), 'created_at': Value('timestamp[s]'), 'versions': {'lean': Value('string'), 'mathlib': Value('string'), 'reap': Value('string'), 'policy_version': Value('string')}})}
because column names don't match
Traceback:    Traceback (most recent call last):
                File "/src/services/worker/src/worker/utils.py", line 147, in get_rows_or_raise
                  return get_rows(
                      dataset=dataset,
                  ...<4 lines>...
                      column_names=column_names,
                  )
                File "/src/libs/libcommon/src/libcommon/utils.py", line 272, in decorator
                  return func(*args, **kwargs)
                File "/src/services/worker/src/worker/utils.py", line 127, in get_rows
                  rows_plus_one = list(itertools.islice(safe_iter(ds, dataset=dataset), rows_max_number + 1))
                File "/src/services/worker/src/worker/utils.py", line 483, in safe_iter
                  yield from ds.decode(False) if ds.features else ds
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2951, in __iter__
                  for key, example in ex_iterable:
                                      ^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2461, in __iter__
                  for key, pa_table in self._iter_arrow():
                                       ~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 2486, in _iter_arrow
                  for key, pa_table in self.ex_iterable._iter_arrow():
                                       ~~~~~~~~~~~~~~~~~~~~~~~~~~~~^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 547, in _iter_arrow
                  for key, pa_table in iterator:
                                       ^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/iterable_dataset.py", line 430, in _iter_arrow
                  for key, pa_table in self.generate_tables_fn(**gen_kwags):
                                       ~~~~~~~~~~~~~~~~~~~~~~~^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 343, in _generate_tables
                  self._cast_table(pa_table, json_field_paths=json_field_paths),
                  ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
                File "/usr/local/lib/python3.14/site-packages/datasets/packaged_modules/json/json.py", line 132, in _cast_table
                  pa_table = table_cast(pa_table, self.info.features.arrow_schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2378, in table_cast
                  return cast_table_to_schema(table, schema)
                File "/usr/local/lib/python3.14/site-packages/datasets/table.py", line 2306, in cast_table_to_schema
                  raise CastError(
                  ...<3 lines>...
                  )
              datasets.table.CastError: Couldn't cast
              schema: string
              wave_id: string
              created_at: timestamp[s]
              store: string
              shard_size: int64
              counts: struct<searches: int64, searches_with_transitions: int64, searches_skipped_unsolved: int64, transiti (... 94 chars omitted)
                child 0, searches: int64
                child 1, searches_with_transitions: int64
                child 2, searches_skipped_unsolved: int64
                child 3, transitions_raw: int64
                child 4, dropped_focus: int64
                child 5, dropped_dedup: int64
                child 6, transitions: int64
                child 7, shards: int64
              dedup: struct<enabled: bool, key: list<item: string>, dropped: int64>
                child 0, enabled: bool
                child 1, key: list<item: string>
                    child 0, item: string
                child 2, dropped: int64
              merge_logprobs: null
              policy_version: string
              policy_versions: list<item: string>
                child 0, item: string
              versions: struct<lean: string, mathlib: string, reap: string, patchset_hash: string>
                child 0, lean: string
                child 1, mathlib: string
                child 2, reap: string
                child 3, patchset_hash: string
              shards: list<item: struct<file: string, count: int64, bytes: int64, sha256: string>>
                child 0, item: struct<file: string, count: int64, bytes: int64, sha256: string>
                    child 0, file: string
                    child 1, count: int64
                    child 2, bytes: int64
                    child 3, sha256: string
              sources: list<item: struct<search: string, problem_id: string, policy_version: string, transitions: int64>>
                child 0, item: struct<search: string, problem_id: string, policy_version: string, transitions: int64>
                    child 0, search: string
                    child 1, problem_id: string
                    child 2, policy_version: string
                    child 3, transitions: int64
              sha256: string
              dataset: string
              schema_version: string
              shard_format: string
              updated_at: timestamp[s]
              batches: list<item: struct<batch_id: string, path: string, manifest: string, shards: list<item: string>, prob (... 187 chars omitted)
                child 0, item: struct<batch_id: string, path: string, manifest: string, shards: list<item: string>, problems: int64 (... 175 chars omitted)
                    child 0, batch_id: string
                    child 1, path: string
                    child 2, manifest: string
                    child 3, shards: list<item: string>
                        child 0, item: string
                    child 4, problems: int64
                    child 5, transitions: int64
                    child 6, solved: int64
                    child 7, terminal_verified: int64
                    child 8, created_at: timestamp[s]
                    child 9, versions: struct<lean: string, mathlib: string, reap: string, policy_version: string>
                        child 0, lean: string
                        child 1, mathlib: string
                        child 2, reap: string
                        child 3, policy_version: string
              license: string
              to
              {'schema_version': Value('string'), 'dataset': Value('string'), 'updated_at': Value('timestamp[s]'), 'shard_format': Value('string'), 'license': Value('string'), 'batches': List({'batch_id': Value('string'), 'path': Value('string'), 'manifest': Value('string'), 'shards': List(Value('string')), 'problems': Value('int64'), 'transitions': Value('int64'), 'solved': Value('int64'), 'terminal_verified': Value('int64'), 'created_at': Value('timestamp[s]'), 'versions': {'lean': Value('string'), 'mathlib': Value('string'), 'reap': Value('string'), 'policy_version': Value('string')}})}
              because column names don't match

Need help to make the dataset viewer work? Make sure to review how to configure the dataset viewer, and open a discussion for direct support.

10-4 Alpha-Proof Trajectories

English · Public trajectory dataset from the 10-4-alpha-proof project: kernel-verified Lean 4 tactic transitions collected by a Python MCTS controller. The first batch (wave_001_fate_m_v001) contains 20 transitions from 20 FATE-M problems, all solved and independently re-verified by the Lean 4 kernel (strict receipt pass).

中文 · 公开轨迹数据集，来自 10-4-alpha-proof 项目：由 Python MCTS 控制器采集、经 Lean 4 内核独立复验的战术转移（transition）。首批 wave_001_fate_m_v001 含 20 题 × 20 条 transition，全部 solved 且通过严格复验收据（strict receipt）。

目录结构 / Layout

catalog.json                        # 批次索引（batch index）
wave_001_fate_m_v001/
  shard-0001.jsonl                  # 20 条 Transition，一行一条（JSON Lines）
  manifest.json                     # 计数 + sha256 + 版本三元组 + policy_version
showcase/
  example-pell.html                 # 示例展示页（Pell 证明：MCTS 会话增量验证可视化）

数据格式 / Schema

每行一个 JSON 对象（Transition，字段与项目训练口径 alphaproof/data/events.py 对齐）：

字段	含义
`prompt`	策略输入（含 Lean 证明状态与生成指令的完整提示）
`action`	选择的战术（tactic）
`value_target`	价值目标（`-d(s)`，d = 剩余步数；终端为 0）
`kind`	`proof` / `disproof` / `timeout`（本批全部为 `proof`）
`solved`	是否求解成功（本批全部 `true`）
`terminal_verified`	终端是否经内核复验（本批全部 `true`）
`logprob_old`	在线支线行为策略 logprob（本批为 `null`，CE 主线不使用）
`reward`	与 `value_target` 分离的奖励位（本批为 `null`）
`extra`	溯源自证：`transition_id`、`problem_id`、`receipt_id`、`attempt_id`、状态/请求/收据 sha256、token ids 等

快速读取：

import json
rows = [json.loads(l) for l in open("wave_001_fate_m_v001/shard-0001.jsonl")]

版本 / Versions

组件	版本
Lean	`4.28.0-rc1`
Mathlib	`v4.28.0-rc1`
Reap	`0090d73c`（完整：`IQuestLab/reap@0090d73c5f739e4d74000e053b00fd0148ff46aa`）
policy_version	`shared-wave-001-input`（behavior_sha256 `68e645ac0427c4ef08633fb249e314f08bac8f285dd4387ac2f21bdfccf85ce9`）
base model	`REAL-Prover-fe76f68d`（sha256 `7ffcbbcea4831ce54254a3514dd792d449587a397bcc3303165cdb611275ff84`）

版本字段均已取证（项目文档 + 生成 receipts 的 behavior_identity），本批没有 unverified: 字段。

来源与可追溯性 / Provenance

实验：fate_m_20x200_ce_vs_online_v2_7b_20261005（CE 支线），run formal-20x10/wave_001
源文件：runs/formal-20x10/wave_001/ce/update/replay/transitions.jsonl
- transitions_sha256 = 688c733e3527692cc1be57b8087a82d5ad8a3b8be25a5a88e7a70d0970b84df7
- 源 replay manifest sha256 = a6bf02a5152572cf6ee71e0f889327cb217aa7990ff2ffb2ccb5d8fd25ffc863
- 源 receipts（ce-receipts-through-wave-001.jsonl）sha256 = 0fe96d85953644150bde6e77e253cb6dff23bd92f7243b4dd89619ae7818266a
生成时间：2026-10-05（UTC）；发布：2026-10-09（UTC）
本目录 shard-0001.jsonl 为 lmc-store 导出格式（canonical）（2026-10-09 对齐）：与源 transitions.jsonl 逐条内容一致、JSON 序列化字节不同；本目录文件 sha256 见 wave_001_fate_m_v001/manifest.json（schema lmc.export_manifest.v1）。
逐文件计数与 sha256 见 wave_001_fate_m_v001/manifest.json

本批 20/20 条来自独立严格复验收据链（verification.result == "verified"，kernel exit 0）。

许可 / License

公开（Public）。 本数据集全部内容公开发布，可自由访问、阅读，并可用于研究与引用。未声明 SPDX 标准许可证；本仓不代表对上游模型权重的再许可。

Public release. All content is published publicly for research use and citation. No SPDX license is declared; this repo does not relicense any upstream model weights.

引用 / Citation

若使用本数据，请引用本数据集（10-4-alpha-proof 项目，alpha-proof-open-source 组织）：

@misc{alpha_proof_104_trajectories,
  title        = {10-4 Alpha-Proof Trajectories},
  author       = {{10-4-alpha-proof project}},
  year         = {2026},
  howpublished = {Hugging Face dataset},
  url          = {https://huggingface.co/datasets/alpha-proof-open-source/10-4-alpha-proof-trajectories}
}

Downloads last month: 14

Total file size:

273 kB