From f327861a00eb689b769447b382ae04561616ddff Mon Sep 17 00:00:00 2001 From: Codex Bot Date: Sun, 8 Nov 2026 03:23:25 +0800 Subject: [PATCH] Hold untitled public assertions and retain explicit withdrawals --- docs/data-sources.md | 2 + paper_atlas/collectors/public_openreview.py | 8 ++-- tests/test_public_openreview.py | 44 ++++++++++++++++++++- 3 files changed, 50 insertions(+), 4 deletions(-) diff --git a/docs/data-sources.md b/docs/data-sources.md index 71b9f62..06b9e17 100644 --- a/docs/data-sources.md +++ b/docs/data-sources.md @@ -93,6 +93,8 @@ paper-atlas audit public-openreview --db /data/paper-index/papers.sqlite \ 输出按会议、届次及 main、position、datasets_benchmarks 分文件。workshop 默认进入单独的 held 清单；有明确 workshop 名称时可通过 `--include-workshops` 单列采集，不合并到主会。Retraction、未映射的轨道或负面标签也保留在 `held-records.jsonl`，不强改为 Reject 或 Withdrawn。当前决定表没有独立 retracted 状态，因此撤回发表记录仍需后续专门建模。 +`parser_version=2` 还会将明确负面标签但标题缺失、空白或 null 的记录放入 `held-records.jsonl`，保留原始状态、forum ID、固定来源版本和行号；不制造标题或导入不可识别的论文，也不让这一条缺失阻断其他可用记录。held 文件哈希与各原因数量保存在 inventory；非文本标题、错误身份、冲突届次和哈希不符仍使整批失败。旧年 ICLR 镜像的明确 `Withdraw` 标签归为 `withdrawn`，`Workshop`、`Poster Workshop`、`Oral Workshop` 均不作为拒稿。采集文件与导入主库仍是两个独立步骤。 + 该来源明确标记为 `curated_public_status_assertion`：保留源 commit、文件哈希、行号、原标签和 forum ID，不声称已核验原始 PC 签名。只保存标题、摘要、公开作者名、关键词、领域、索引与决定来源，不缓存镜像中的个人资料、邮箱、性别、评审内容或 PDF。所有输出清单均为 partial；盘点中的零条记录仅描述这个固定镜像。 入库前审计核验文件哈希、数量、轨道和来源声明，并用真实 forum ID 对比同届同轨道已有投稿。存在不同决定时阻止整批导入，交由来源和时序核查；另一会议的后续录用属于独立投稿历史，不构成该次拒稿的冲突。同一 forum 的镜像不新增投稿事件，但论文实体的跨平台去重仍须在全量结束后核验。审计通过的报告包含带 SHA-256 的导入队列；应在单写入维护锁下及时串行导入，主库变化后重新审计。 diff --git a/paper_atlas/collectors/public_openreview.py b/paper_atlas/collectors/public_openreview.py index bde6ab7..56292a4 100644 --- a/paper_atlas/collectors/public_openreview.py +++ b/paper_atlas/collectors/public_openreview.py @@ -27,7 +27,7 @@ RAW_ROOT = 'https://raw.githubusercontent.com/papercopilot/openreview/' + REVISI TREE_URL = 'https://api.github.com/repos/papercopilot/openreview/git/trees/' + REVISION + '?recursive=1' NEGATIVE = {'reject': 'rejected', 'rejected': 'rejected', 'desk reject': 'desk_rejected', 'desk rejected': 'desk_rejected', - 'withdrawn': 'withdrawn', 'withdrawal': 'withdrawn'} + 'withdraw': 'withdrawn', 'withdrawn': 'withdrawn', 'withdrawal': 'withdrawn'} TRACKS = {'main': 'main', 'Position': 'position', 'Datasets & Benchmarks': 'datasets_benchmarks'} WORKSHOP_REJECT_LABELS = {"aim-fm workshop @ neurips'24 reject", @@ -90,7 +90,9 @@ def normalize(row, family, year, provenance, include_workshops=False): return None, 'unmapped_venue' title = row.get('title') abstract = row.get('abstract') - if not isinstance(title, str) or not title.strip(): + if title is None or (isinstance(title, str) and not title.strip()): + return None, 'missing_title' + if not isinstance(title, str): raise ValueError('Public submission lacks a title') if abstract is not None and not isinstance(abstract, str): raise ValueError('Abstract must be text or null') @@ -215,7 +217,7 @@ def _collect_files(out, selected, include_workshops, fetcher): name = f'public-openreview-{family}-{year}-{track}.jsonl' path = stage / name path.write_text(''.join(json.dumps(row, ensure_ascii=False) + '\n' for row in records)) - manifest = {'source': SOURCE, 'parser_version': 1, 'venue': records[0]['venue'], + manifest = {'source': SOURCE, 'parser_version': 2, 'venue': records[0]['venue'], 'year': year, 'track': track, 'count': len(records), 'status': 'partial', 'file': str(out / name), 'file_sha256': hashlib.sha256(path.read_bytes()).hexdigest(), 'list_url': REPOSITORY + '/tree/' + REVISION, 'source_revision': REVISION, diff --git a/tests/test_public_openreview.py b/tests/test_public_openreview.py index 6471ab3..03412ae 100644 --- a/tests/test_public_openreview.py +++ b/tests/test_public_openreview.py @@ -45,10 +45,12 @@ class PublicOpenReviewTests(unittest.TestCase): def test_only_explicit_decisions_and_verified_scope_become_events(self): for value in ('Pending', 'Active', 'Submitted to NeurIPS 2025', 'TBD', '', - 'Low score', 'ICML 2025 Retracted by Authors', 'Do not reject'): + 'Low score', 'ICML 2025 Retracted by Authors', 'Do not reject', + 'Workshop', 'Poster Workshop', 'Oral Workshop'): self.assertIsNone(negative_status(value), value) self.assertEqual(negative_status('Reject'), 'rejected') self.assertEqual(negative_status('Desk rejected'), 'desk_rejected') + self.assertEqual(negative_status('Withdraw'), 'withdrawn') self.assertEqual(negative_status('NeurIPS 2023 Conference Withdrawn Submission'), 'withdrawn') for raw, expected in [('main', 'main'), ('Position', 'position'), ('Datasets & Benchmarks', 'datasets_benchmarks')]: @@ -98,6 +100,46 @@ class PublicOpenReviewTests(unittest.TestCase): self.assertEqual(list(Path(tmp).glob('*.jsonl')), []) self.assertFalse((Path(tmp) / 'public-openreview-inventory.json').exists()) + def test_missing_title_is_held_without_losing_valid_negative_records(self): + with tempfile.TemporaryDirectory() as tmp: + result = self.collect(tmp, {'venues/iclr/iclr2017.jsonl': [ + self.row(id='SJUdkecgx', title=''), + self.row(id='NoTitle001', title=None), + self.row(id='Whitespace', title=' '), + self.row(id='ValidReject'), + self.row(id='Withdraw01', status='Withdraw'), + self.row(id='Workshop01', status='Workshop'), + ]}, families=['iclr'], years=[2017]) + self.assertEqual(result['input_rows'], 6) + self.assertEqual(result['normalized_records'], 2) + self.assertEqual(result['source_files'][0]['excluded_or_held'], + {'missing_title': 3, 'not_explicit_negative': 1}) + held_path = Path(result['held_file']) + self.assertEqual(hashlib.sha256(held_path.read_bytes()).hexdigest(), + result['held_file_sha256']) + held = [json.loads(line) for line in held_path.read_text().splitlines()] + self.assertEqual([row['id'] for row in held], ['SJUdkecgx', 'NoTitle001', 'Whitespace']) + self.assertTrue(all(row['reason'] == 'missing_title' and row['status'] == 'Reject' + for row in held)) + self.assertEqual([row['provenance']['source_line'] for row in held], [1, 2, 3]) + self.assertNotIn('private@example.org', held_path.read_text()) + self.assertNotIn('private review', held_path.read_text()) + manifest = result['files'][0] + self.assertEqual((manifest['parser_version'], manifest['status']), (2, 'partial')) + rows = [json.loads(line) for line in Path(manifest['file']).read_text().splitlines()] + self.assertEqual([(row['identifiers']['openreview'], row['status']) for row in rows], + [('ValidReject', 'rejected'), ('Withdraw01', 'withdrawn')]) + for row in held: + self.assertEqual(row['provenance']['source_revision'], REVISION) + + def test_nontext_title_still_fails_collection(self): + with tempfile.TemporaryDirectory() as tmp: + with self.assertRaisesRegex(ValueError, 'lacks a title'): + self.collect(tmp, {'venues/iclr/iclr2017.jsonl': [self.row(title=['invalid'])]}, + families=['iclr'], years=[2017]) + self.assertFalse((Path(tmp) / 'public-openreview-inventory.json').exists()) + self.assertEqual(list(Path(tmp).glob('*.jsonl')), []) + def test_mirror_reimport_preserves_one_attempt_and_later_acceptance_elsewhere(self): with tempfile.TemporaryDirectory() as tmp: root = Path(tmp)