diff --git a/.gitea/workflows/check.yml b/.gitea/workflows/check.yml index 34490c0..ccfa4bb 100644 --- a/.gitea/workflows/check.yml +++ b/.gitea/workflows/check.yml @@ -75,11 +75,15 @@ jobs: path.write_bytes(body.encode('utf-8')) subprocess.run([python,'-I','-B','-m','paper_atlas','collect','mm2026-publications','--registry-snapshots',str(mm_pub_registry),'--program-snapshots',str(mm_snapshots),'--parent-snapshot',str(mm_pub_registry/'parent.json'),'--out',str(root/'installed-mm2026-publications')],cwd=outside,check=True) mm_pub_report=json.loads((root/'installed-mm2026-publications/collection-report.json').read_text()) - assert (mm_pub_report['records'],mm_pub_report['unique_registered_dois'],mm_pub_report['abstracts_from_exact_program_correspondence'],mm_pub_report['output_missing_abstract_count'])==(1918,1918,1717,201) - assert mm_pub_report['unmatched_program_count']==417 and not mm_pub_report['database_written'] and mm_pub_report['approved_for_import']==0 + assert (mm_pub_report['records'],mm_pub_report['unique_registered_dois'],mm_pub_report['abstracts_from_exact_program_correspondence'],mm_pub_report['output_missing_abstract_count'])==(1918,1918,1743,175) + assert mm_pub_report['unmatched_program_count']==391 and not mm_pub_report['database_written'] and mm_pub_report['approved_for_import']==0 mm_pub_rows={r['external_id']:r for f in mm_pub_report['files'] for r in map(json.loads,pathlib.Path(f['file']).read_text().splitlines())} assert len(mm_pub_rows)==1918 - for gold in mm_pub_fixture['independent_gold']: + mm_format_gold=json.loads((src/'tests/fixtures/mm2026-title-formatting-review.json').read_text()) + mm_format_overrides={g['doi']:g for g in [*mm_format_gold['gold'],mm_format_gold['held_gold']]} + assert mm_pub_report['abstracts_from_reviewed_title_formatting']==26 and mm_pub_report['abstracts_from_literal_program_correspondence']==1717 + for original_gold in mm_pub_fixture['independent_gold']: + gold=mm_format_overrides.get(original_gold['doi'],original_gold) row=mm_pub_rows[gold['doi']] assert (row['title'],row['authors'],row['track'])==(gold['title'],gold['authors'],gold['track']) assert (hashlib.sha256(row['abstract'].encode()).hexdigest() if row['abstract'] is not None else None)==gold['abstract_sha256'] diff --git a/README.md b/README.md index b7ce3cd..daabd89 100644 --- a/README.md +++ b/README.md @@ -145,4 +145,4 @@ ECCV 2026 的[暂定节目采集器](docs/eccv2026-program.md)合并重复节目 `collect mm2026-program` 从已冻结官方快照恢复全部 28 个来源 track 的 2,143 个节目 ID,保留 4,326 次展示的来源位置与场次上下文,单列 216 个摘要缺失项;输出供贡献类型、出版及作品身份核查使用,不是可直接导入的论文/决定队列。见 [MM 2026 使用、来源和恢复说明](docs/mm2026-source-status.md)。 -`collect mm2026-publications` 核验完整论文集的 1,918 个注册出版 DOI,并单独关联节目分轨、展示上下文和摘要。1,717 份节目摘要明确标注来源,192 项标题/作者版本差异保留为待核;所有节目及未匹配项均保留。注册文章类型不等于研究类型,输出尚需数据库身份/历史审查。见 [MM 2026 出版条目采集与恢复](docs/mm2026-publications.md)。 +`collect mm2026-publications` 核验完整论文集的 1,918 个注册出版 DOI,并单独关联节目分轨、展示上下文和摘要。1,743 份节目摘要明确标注来源(其中 26 份来自已复核的标题排版对应),166 项标题/作者版本差异保留为待核;所有节目及未匹配项均保留。注册文章类型不等于研究类型,输出尚需数据库身份/历史审查。见 [MM 2026 出版条目采集与恢复](docs/mm2026-publications.md)。 diff --git a/docs/mm2026-publications.md b/docs/mm2026-publications.md index 374d114..f5a0cf2 100644 --- a/docs/mm2026-publications.md +++ b/docs/mm2026-publications.md @@ -11,7 +11,7 @@ paper-atlas collect mm2026-publications \ --registry-snapshots /data/paper-index/collected \ --program-snapshots /data/paper-index/collected \ --parent-snapshot /data/paper-index/collected/mm2026-publication-metadata-v1/parent.json \ - --out /data/paper-index/collected/mm2026-publications-native-v1 + --out /data/paper-index/collected/mm2026-publications-native-v3 ``` 输出目录必须不存在。失败输出保留 `failed.json`,完整 `collection-report.json` 只在全部校验成功后原子生成;不要覆盖输出或因旧进度文件、日志安静重跑。先观察实际单元/报告,再选择新的恢复目录。 @@ -28,27 +28,31 @@ ISBN查询和父Unixref XML只返回父项,不能代表逐篇目录。宽泛 ` ## 分轨、摘要与版本差异 -按原始题名的Unicode/字母数字规范化与完整作者多重集合比对,1726项有唯一节目对应。源track映射到对应输出分区,并保留原始标签、节目ID、机构和所有展示上下文;不生成全局人物别名。其余122项作者版本差异、70项题名版本差异保留在 `proceedings_unclassified`,不强配分轨或静默排除。 +按原始题名的 Unicode/字母数字规范化与完整作者多重集合比对,1726 项有唯一节目对应。原始题名没有候选时,再比较经过受限排版投影的题名:只移除平衡、无属性的 `i`、`scp`、`sup`、`sub` 标签,保留上下标内容,HTML 实体只解码一次。未知或不平衡标签、属性、注释及声明不参与这种投影;不使用模糊题名或人物别名。 + +26 项新增对应均保留完整作者多重集合一致、唯一节目 ID 以及逐项复核名单,复制节目原文的摘要、track 和展示上下文。原始注册题名、作者和节目字段继续逐字保留,比较文本单独存入 `program_title_formatting_comparison`;这只证明节目材料对应,不批准科学作品合并。原始题名已有作者分歧时,不通过排版回退绕过。其余 122 项原始作者分歧、1 项排版题名下的作者分歧、43 项题名差异,共 166 项仍在 `proceedings_unclassified`。 | 输出分区 | 注册条目 | | --- | ---: | -| main | 1533 | +| main | 1555 | | grand_challenge | 77 | | interactive_art | 17 | -| datasets | 40 | +| datasets | 41 | | reproducibility | 1 | | bni | 22 | | open_source | 11 | | doctoral_symposium | 2 | -| demo_and_video | 20 | +| demo_and_video | 23 | | industry_talk | 3 | -| proceedings_unclassified | 192 | +| proceedings_unclassified | 166 | -注册摘要1918项全缺失。1717项有明确对应的节目摘要,逐字保留并标明 `abstract_source=official_program_exact_metadata_correspondence`;注册原始摘要仍标为缺失,绝不伪称由出版者提供。其余201项保持缺失,包括9项已对应节目本身没有摘要的条目及192个版本待核条目。没有用模糊匹配把待核摘要强加给论文。 +注册摘要 1918 项仍全缺失。1743 项有明确对应的节目摘要:1717 项使用 `abstract_source=official_program_exact_metadata_correspondence`;新增 26 项使用 `official_program_reviewed_title_formatting_correspondence`,两者都逐字复制节目原文。报告分别列出 `abstracts_from_literal_program_correspondence` 和 `abstracts_from_reviewed_title_formatting`;保留的汇总字段 `abstracts_from_exact_program_correspondence` 为 1743,表示两类非模糊对应之和。 -8个作者仅有单一姓名字段,保留登记值并标记,不猜测given/family;37个含排版标记或实体的题名保留字面原文并标记,不能损失数学上下标或假定排版版本已经解决。 +其余 175 项保持摘要缺失,包括 9 项节目本身没有摘要及 166 个版本待核条目。注册原始摘要仍标为缺失,不能把节目内容说成出版者摘要。 -完整2143个节目ID和4326个展示位置仍在匹配证据或417个未匹配节目中,28个原始track均保留。417个节目与192个出版版本待核来自两种观察口径,不能直接相加当成不同论文。Workshop、艺术、行业等贡献不能因为不在这个论文集范围就被当成拒稿、撤稿或不值得保留。 +8 个作者仅有单一姓名字段,保留登记值并标记,不猜测 given/family;37 个含排版标记或实体的题名仍保留字面原文和原始标记。排版比较保留上下标字符,不改写原始字段。 + +完整2143个节目ID和4326个展示位置仍在匹配证据或391个未匹配节目中,28个原始track均保留。391个节目与166个出版版本待核来自两种观察口径,不能直接相加当成不同论文。Workshop、艺术、行业等贡献不能因为不在这个论文集范围就被当成拒稿、撤稿或不值得保留。 注册出版支持来源侧 `status=accepted` 的发表观察,`raw_decision` 明确不是原始签署决定;出版日期仅放在注册元数据中,不伪造决定日期。`mm_registry_program:crossref:type` 与 `mm_registry_program:mm_program:track` 分别保存注册书目类型和节目 track;共同的派生来源前缀满足写入器的来源校验,两个标签命名空间仍相互独立,原始平台字段与证据继续保留在各自元数据中,`research_type_classified=false`:书目类型不是研究类型,行业、艺术、demo条目不能默认变成主轨研究论文。 @@ -64,4 +68,6 @@ ISBN查询和父Unixref XML只返回父项,不能代表逐篇目录。宽泛 ` 节目原始来源恢复包见 [MM来源说明](mm2026-source-status.md)。本地独立核验备份位于 `/Users/zzyang/Documents/Codex/2026-11-06/paper-index/`。固定回执/正文哈希变化、总量变化、未知注册上下文或节目对应改变都须重新核查,不能放宽为无条件接受。最终全库去重、近似题名作者、标识冲突、摘要、约束、覆盖与查询审计仍是独立必做阶段。 -实际导入回归在隔离的 SQLite 和 PostgreSQL 测试库中运行全部1918条、全部11个输出分区,核验原始字段、1918个书目类型标签、1726个节目标签、publication_doi 身份、partial覆盖,以及已有作品的规范信息和旧拒稿决定不被覆盖。这些测试不会写入生产 research schema;测试通过仍不替代逐条生产身份审阅。 +实际导入回归在隔离的 SQLite 和 PostgreSQL 测试库中运行全部1918条、全部11个输出分区,核验原始字段、1918个书目类型标签、1752个节目标签、publication_doi 身份、partial覆盖,以及已有作品的规范信息和旧拒稿决定不被覆盖。这些测试不会写入生产 research schema;测试通过仍不替代逐条生产身份审阅。 + +旧的 `native-v1`、`native-v2` 来源目录及已消费的导入队列不可覆盖或重放。新版本写入新的输出目录,先核对已导入记录的源字段和分轨是否变化;新增来源关联仍须新鲜身份预检、保护基线和独立验收。程序输出的分区预期数量已改变,未来增量 manifest 必须明确保留旧覆盖父值并声明新预期值,不能重置已导入计数。 diff --git a/paper_atlas/collectors/mm_publications2026.py b/paper_atlas/collectors/mm_publications2026.py index 22d8393..c3f341e 100644 --- a/paper_atlas/collectors/mm_publications2026.py +++ b/paper_atlas/collectors/mm_publications2026.py @@ -7,6 +7,7 @@ import argparse from collections import Counter, defaultdict from datetime import datetime import hashlib +from html.parser import HTMLParser import json from pathlib import Path import re @@ -115,31 +116,138 @@ def load_registry(root, parent_path): return evidence,rows -def correlate(rows, observations): - titles=defaultdict(list) - for row in observations:titles[key(row['title'])].append(row) + +class _TitleFormatting(HTMLParser): + """Decode text once, removing only balanced, attribute-free reviewed tags.""" + def __init__(self): + super().__init__(convert_charrefs=True) + self.parts=[];self.stack=[];self.valid=True + + def handle_data(self, text): + self.parts.append(text) + + def handle_starttag(self, tag, attrs): + if tag not in {'i','scp','sup','sub'} or attrs: + self.valid=False + else: + self.stack.append(tag) + + def handle_endtag(self, tag): + if not self.stack or self.stack[-1]!=tag: + self.valid=False + else: + self.stack.pop() + + def handle_startendtag(self, tag, attrs): + self.valid=False + + def handle_comment(self, text): + self.valid=False + + def handle_decl(self, text): + self.valid=False + + def handle_pi(self, text): + self.valid=False + + def unknown_decl(self, text): + self.valid=False + + +def formatting_text(title): + # HTMLParser may silently discard a trailing incomplete start tag. + # Check lexical coverage before parsing so no title content can disappear. + literal=re.sub(r'','',title,flags=re.I) + if '<' in literal or '>' in literal: + return None + parser=_TitleFormatting() + parser.feed(title);parser.close() + text=''.join(parser.parts) + return text if parser.valid and not parser.stack and text.strip() and '<' not in text and '>' not in text else None + + +FORMATTING_MATCHES = {'10.1145/3767308.3834907': 'mfp0053', + '10.1145/3767308.3834936': 'mfp0229', + '10.1145/3767308.3834946': 'mfp0297', + '10.1145/3767308.3834956': 'mfp0339', + '10.1145/3767308.3835035': 'mfp0798', + '10.1145/3767308.3835055': 'mfp0903', + '10.1145/3767308.3835101': 'mfp1175', + '10.1145/3767308.3835374': 'mfp2617', + '10.1145/3767308.3835424': 'mfp2909', + '10.1145/3767308.3835452': 'mfp3048', + '10.1145/3767308.3835460': 'mfp3109', + '10.1145/3767308.3835544': 'mfp3554', + '10.1145/3767308.3835591': 'mfp3783', + '10.1145/3767308.3835667': 'mfp4173', + '10.1145/3767308.3835763': 'mfp4778', + '10.1145/3767308.3835895': 'mfp5537', + '10.1145/3767308.3835959': 'mfp5823', + '10.1145/3767308.3836215': 'mfp7332', + '10.1145/3767308.3836272': 'mfp7635', + '10.1145/3767308.3836293': 'mfp7732', + '10.1145/3767308.3836447': 'mfp8648', + '10.1145/3767308.3836600': 'mfp9815', + '10.1145/3767308.3838656': 'mds177', + '10.1145/3767308.3839012': 'mde41', + '10.1145/3767308.3839024': 'mde15', + '10.1145/3767308.3839028': 'mde02'} + + +def match_program(rows, observations): + titles=defaultdict(list);formatted=defaultdict(list) + for p in observations: + titles[key(p['title'])].append(p) + text=formatting_text(p['title']) + if text is not None:formatted[key(text)].append(p) matched=set();kinds=Counter() for row in rows: + row.pop('program',None);row.pop('formatting_comparison',None) candidates=titles[key(row['title'])] + method='literal_title' + # A literal-title author disagreement must not be bypassed by fallback. + if not candidates: + text=formatting_text(row['title']) + if text is not None: + candidates=formatted[key(text)] + if candidates:method='reviewed_title_formatting' exact=[p for p in candidates if author_names(p['authors'])==author_names(row['authors'])] if len(exact)==1: - row['kind']='unique_program_title_fullauthor_agreement_not_identity_approval' + row['kind']=('unique_program_format_title_fullauthor_agreement_not_identity_approval' + if method=='reviewed_title_formatting' else 'unique_program_title_fullauthor_agreement_not_identity_approval') row['program']=exact[0];matched.add(exact[0]['paper_id']) if exact[0]['track'] not in TRACKS: raise ValueError('New registered program track requires review') elif len(exact)>1:row['kind']='multiple_program_ids_same_title_authors_hold' - elif candidates:row['kind']='program_author_version_requires_review' + elif candidates: + row['kind']=('program_format_author_version_requires_review' + if method=='reviewed_title_formatting' else 'program_author_version_requires_review') else:row['kind']='no_exact_program_title_requires_review' row['program_title_candidates']=[p['paper_id'] for p in candidates] row['program_exact_candidates']=[p['paper_id'] for p in exact] + if method=='reviewed_title_formatting': + row['formatting_comparison']={'method':'balanced_attribute_free_i_scp_sup_sub_and_single_entity_decode_then_nfkd_alnum', + 'registry_comparison_text':formatting_text(row['title']), + 'program_comparison_texts':{p['paper_id']:formatting_text(p['title']) for p in candidates}, + 'raw_titles_preserved':True,'work_identity_approved':False} kinds[row['kind']]+=1 - expected={'unique_program_title_fullauthor_agreement_not_identity_approval':1726, - 'program_author_version_requires_review':122,'no_exact_program_title_requires_review':70} - if kinds != expected: - raise ValueError('Reviewed program/publication correspondence changed') return dict(kinds),[p for p in observations if p['paper_id'] not in matched] +def correlate(rows, observations): + kinds,unmatched=match_program(rows,observations) + expected={'unique_program_title_fullauthor_agreement_not_identity_approval':1726, + 'unique_program_format_title_fullauthor_agreement_not_identity_approval':26, + 'program_author_version_requires_review':122, + 'program_format_author_version_requires_review':1, + 'no_exact_program_title_requires_review':43} + actual={r['doi']:r['program']['paper_id'] for r in rows + if r['kind']=='unique_program_format_title_fullauthor_agreement_not_identity_approval'} + if kinds != expected or actual != FORMATTING_MATCHES: + raise ValueError('Reviewed program/publication correspondence changed') + return kinds,unmatched + + def collect(out, *, registry_snapshots, program_snapshots, parent_snapshot): out=Path(out) if out.exists() or out.is_symlink(): @@ -170,7 +278,7 @@ def collect(out, *, registry_snapshots, program_snapshots, parent_snapshot): 'venue':'MM','year':2026,'track':track,'status':'accepted', 'raw_decision':'Registered publication in MM2026 proceedings; not an original signed decision.', 'evidence_url':receipt['url'],'observed_at':receipt['at'],'tags':tags, - 'metadata':{'registry_fields':item['fields'],'registry_original_field_names':item['original_field_names'], + 'metadata':{**({'program_title_formatting_comparison':item['formatting_comparison']} if 'formatting_comparison' in item else {}),'registry_fields':item['fields'],'registry_original_field_names':item['original_field_names'], 'registry_source_evidence':{'file':'evidence/registry/'+receipt['file'],'sha256':receipt['sha256'], 'url':receipt['url'],'json_pointer':'/message/items/'+str(item['source_position']), 'canonical_object_json_sha256':item['original_canonical_json_sha256']}, @@ -180,7 +288,7 @@ def collect(out, *, registry_snapshots, program_snapshots, parent_snapshot): 'program_fields':p,'program_match_kind':item['kind'], 'program_title_candidates':item['program_title_candidates'], 'program_exact_candidates':item['program_exact_candidates'], - 'abstract_source':'official_program_exact_metadata_correspondence' if abstract is not None else 'not_available_from_reviewed_sources', + 'abstract_source':('official_program_reviewed_title_formatting_correspondence' if item.get('formatting_comparison') else 'official_program_exact_metadata_correspondence') if abstract is not None else 'not_available_from_reviewed_sources', 'registry_abstract_available':False,'original_decision_signature_verified':False, 'scientific_work_identity_approved':False,'research_type_classified':False, 'review_flags':item['flags'], @@ -189,7 +297,7 @@ def collect(out, *, registry_snapshots, program_snapshots, parent_snapshot): for track,rows in sorted(buckets.items()): path=out/f'mm-publications-2026-{track}.jsonl' raw=''.join(json.dumps(row,ensure_ascii=False)+'\n' for row in rows).encode();path.write_bytes(raw) - manifest={'source':SOURCE,'parser_version':1,'venue':'MM','year':2026,'track':track,'count':len(rows), + manifest={'source':SOURCE,'parser_version':2,'venue':'MM','year':2026,'track':track,'count':len(rows), 'file':str(path),'file_sha256':digest(raw),'list_url':'https://doi.org/'+PARENT_DOI, 'status':'partial','missing_abstract_count':sum(row['abstract'] is None for row in rows), 'scope':'Registered container partition only; not allMMprogram/workshops, independent works or public-negative coverage.'} @@ -199,12 +307,13 @@ def collect(out, *, registry_snapshots, program_snapshots, parent_snapshot): 'matched_program_track_counts':dict(known_tracks),'unmatched_program':unmatched, 'unmatched_program_count':len(unmatched),'registry_missing_abstract_count':1918, 'abstracts_from_exact_program_correspondence':copied,'output_missing_abstract_count':missing, + 'abstracts_from_literal_program_correspondence':copied-26,'abstracts_from_reviewed_title_formatting':26, 'registered_single_name_fields':sum(f['kind']=='single_registered_name_field_not_split' for r in publications for f in r['flags']), 'registered_title_markup_flags':sum(any(f['kind']=='literal_title_markup_or_entity_retained_requires_review' for f in r['flags']) for r in publications), 'database_written':False,'approved_for_import':0,'network_requests':0,'pdfs_downloaded':False, 'public_negative_availability':'unknown_no_negative_decisions_inferred', - 'scope':'1918registeredpublication DOIs, not1918proven distinct scientificworks.192title/authorvariants remainunclassified,417unmatchedprogram items retained. All28programtracks remain independent census; article type never implies research. Per-work/source/DOI/history review andfinalwholeDBaudit remain.'} - if (sum(f['count'] for f in outputs)!=1918 or copied!=1717 or missing!=201 or len(unmatched)!=417): + 'scope':'1918registeredpublication DOIs, not1918proven distinct scientificworks.166title/authorvariants remainunclassified,391unmatchedprogram items retained.26formattingcorrespondences retainedseparately;rawtitlesunchanged. All28programtracks remain independent census; article type never implies research. Per-work/source/DOI/history review andfinalwholeDBaudit remain.'} + if (sum(f['count'] for f in outputs)!=1918 or copied!=1743 or missing!=175 or len(unmatched)!=391): raise ValueError('Reviewed output partition or abstract census changed') part=out/'collection-report.json.part';part.write_text(json.dumps(report,ensure_ascii=False,indent=2)+'\n');part.replace(out/'collection-report.json') return report diff --git a/tests/fixtures/mm2026-title-formatting-review.json b/tests/fixtures/mm2026-title-formatting-review.json new file mode 100644 index 0000000..8ba842a --- /dev/null +++ b/tests/fixtures/mm2026-title-formatting-review.json @@ -0,0 +1,706 @@ +{ + "status": "reviewed26_presentation_only_title_correspondences_not_work_identity_approval", + "gold": [ + { + "doi": "10.1145/3767308.3834907", + "title": "H\n 2\n AL: Hyperbolic Hierarchy-aware Aggregative Learning for Registration-based Few-shot Medical Image Segmentation", + "authors": [ + "Jia Wang", + "Jiaming Cai", + "Zunying Hu", + "Zhanjie Wu", + "Jinyuan Liu", + "Hua Cheng", + "Yun Peng", + "Xin Fan" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-004.json", + "source_position": 246, + "registry_object_sha256": "dd4a627c13ca806e15e59d3ae894c9216d14d33c85bc5ce7362b51fad8c8a692", + "program_id": "mfp0053", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "516fac88f1b7a2821c1e2b97cc41530f32abbe2514b65b5823427ff3ecdc9a23", + "program_context_pointers": [ + "/entries/129/papers/19", + "/entries/129/subsessions/1/items/7" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3834936", + "title": "M\n 3\n TR: Temporal Retrieval Enhanced Multi-Modal Micro-video Popularity Prediction", + "authors": [ + "Jiacheng Lu", + "Weijian Wang", + "Mingyuan Xiao", + "Yang Hua", + "Tao Song", + "Bo Peng", + "Cheng Hua", + "Haibing Guan" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-005.json", + "source_position": 25, + "registry_object_sha256": "93340f8b923d7cc0d848941107351fb661d95effa35b1869921e7163da36bae4", + "program_id": "mfp0229", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "6968e6a2d364075020e0b6c95c9076742531681232c40df511a2e8992c6b25f9", + "program_context_pointers": [ + "/entries/109/papers/27", + "/entries/109/subsessions/2/items/3" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3834946", + "title": "PAST:\n Prompt-Adaptive Sampling Termination for Efficient Diffusion Model", + "authors": [ + "Renye Yan", + "Jikang Cheng", + "You Wu", + "Wei Peng", + "Zongwei Wang", + "Ling Liang", + "Yimao Cai" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-005.json", + "source_position": 35, + "registry_object_sha256": "4210191c2871090711820aef9aff06289694380c9d0194ec9b8b1e5e40fad00d", + "program_id": "mfp0297", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "b47bd7502bf3a8e9e9dc87e3d9552efe928cc455f843d527a4860c430c5e9490", + "program_context_pointers": [ + "/entries/61/papers/40", + "/entries/61/subsessions/3/items/4" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3834956", + "title": "SwiftExplorer:\n Training-free diffusion model alignment with Swift Diversity Exploration", + "authors": [ + "Renye Yan", + "Jikang Cheng", + "You Wu", + "Baojin Huang", + "Wei Peng", + "Zongwei Wang", + "Ling Liang", + "Yimao Cai" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-006.json", + "source_position": 235, + "registry_object_sha256": "7f2812ca4072437ecaaa585e85139e7cba0e0d721f9a2c0ac4e9897739faed47", + "program_id": "mfp0339", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "f7809f57290e44fe291ffdc07ea956e5484aa491390cd385db75d056f95b840a", + "program_context_pointers": [ + "/entries/88/papers/48", + "/entries/88/subsessions/4/items/0" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835035", + "title": "Balancing Efficiency and Efficacy:\n Training-Free Attention-Guided Switching Between Explicit and Latent Thoughts for MLLMs", + "authors": [ + "Haoqian Kang", + "Liupeng Li", + "Kuofeng Gao", + "Jinpeng Wang", + "Zhenyu Lu", + "Bin Chen", + "Ke Chen", + "Yaowei Wang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-004.json", + "source_position": 8, + "registry_object_sha256": "3d76f692754dc159642eeb1706491631267d437a95c4459674361ebf35f416b9", + "program_id": "mfp0798", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "477e42fa3d5c8236d6415ff18ba26d799d659b9e2f3f09c1d80bc0971dbca125", + "program_context_pointers": [ + "/entries/88/papers/39", + "/entries/88/subsessions/3/items/3" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835055", + "title": "TM\n 2\n Q-Det: Boosting Video Object Detection via Text-Guided Spatiotemporal Mamba and Multimodal Query Learning", + "authors": [ + "Qiang Qi", + "Wenqi Shang", + "Xiao Wang", + "Shuyuan Lin" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-004.json", + "source_position": 28, + "registry_object_sha256": "7e3c3e6acdad588970c681bdb92969f5134086baa2a5a7b9414871135d5ea137", + "program_id": "mfp0903", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "8549d38e781896079ab04ef6a771a4061a0fe42df2fa5c5c03a037d59cc9f548", + "program_context_pointers": [ + "/entries/129/papers/55", + "/entries/129/subsessions/4/items/7" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835101", + "title": "ConfAvatar:\n Confidence-Aware Self-Bootstrapping for Monocular 4D Head Avatar Reconstruction", + "authors": [ + "Zifan Zheng", + "Yu Luo", + "Yuxin Huang", + "Chuancheng Shi", + "Mohammad Masbaul Alam Polash", + "Zhiyong Wang", + "Kun Hu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-008.json", + "source_position": 82, + "registry_object_sha256": "b838756bd0b6af036e408936f8f13a6398d83e524ce0282167199eca092bc7d8", + "program_id": "mfp1175", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "cb8dbe67e16e163e5b136ad3401fc57c827905e9704d7c0a84961bc8f12565f4", + "program_context_pointers": [ + "/entries/109/papers/86", + "/entries/109/subsessions/7/items/2" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835374", + "title": "DateGBC:\n Deep Fuzzy Granular-Ball Clustering via Adaptive Topology Evolution for Unsupervised Image Segmentation", + "authors": [ + "Qianpeng Chong", + "Jiakun Wen", + "Guangyi Wei", + "Wenyi Zeng" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-006.json", + "source_position": 59, + "registry_object_sha256": "a7fe694a04449e841be5181888ae35786d0a4938d87eca1fffb0776d4d350e30", + "program_id": "mfp2617", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "b57b7026c750aeebb47598860b82980215c9e4e4935f6e8611a31b2fe64f73b9", + "program_context_pointers": [ + "/entries/61/papers/80", + "/entries/61/subsessions/6/items/8" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835424", + "title": "RTCBench\n : A Scalable Benchmarking Framework for Multi-Party Video Conferencing", + "authors": [ + "Evan Ram", + "Tuan Tran", + "Youngwook Son", + "Sangtae Ha" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-005.json", + "source_position": 108, + "registry_object_sha256": "b21198e0eba35d175bc83292be8c87bca2b80a9c8e01f40b640a4e9efc7ecde5", + "program_id": "mfp2909", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "2398ec6539c50b14c7ed4fa773277cc860151ab77b5430b49d935f88e90a487e", + "program_context_pointers": [ + "/entries/129/papers/130", + "/entries/129/subsessions/8/items/22" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835452", + "title": "HeadRouter:\n Dynamic Head-Weight Routing for Task-Adaptive Audio Token Pruning in Large Audio Language Models", + "authors": [ + "He Peize", + "Yaodi Luo", + "Xiaoqian Liu", + "Xuyang Liu", + "Jiahang Deng", + "Yaosong Du", + "Li Bangyu", + "Xiyan Gui", + "Yuxuan Chen", + "Linfeng Zhang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-004.json", + "source_position": 167, + "registry_object_sha256": "ee31b12972ea5844a50c6c63a6fc3430875c3388a4bf261f8e76cab31112d351", + "program_id": "mfp3048", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "c455de7c27f43d0fb49990ead5fca795334c0b613eafa2b3c6150eb147d9c3fa", + "program_context_pointers": [ + "/entries/46/papers/78", + "/entries/46/subsessions/6/items/6" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835460", + "title": "CoAnchor:\n Robust Collaborative Perception under Spatio-Temporal Misalignment via Object-Level Anchors", + "authors": [ + "Chi Li", + "Rui Lin", + "Aobo Ji", + "Dongzhu Xu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-004.json", + "source_position": 175, + "registry_object_sha256": "7c41b0dcb65ad27a1f0961f095889ccc73044776250ec816cf95fc25c30da7f1", + "program_id": "mfp3109", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "5f3d97775d83528d6f41fb2fc481d1848cac7c3f9de1d469039450f9b2ecbff1", + "program_context_pointers": [ + "/entries/109/papers/121", + "/entries/109/subsessions/8/items/13" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835544", + "title": "OC-GS:\n Object-Centric Representation and Demand Alignment for Immersive 3D Gaussian Streaming", + "authors": [ + "Boyan Li", + "Ruijun Mao", + "Zhicheng Liang", + "Dayou Zhang", + "Fangxin Wang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-006.json", + "source_position": 129, + "registry_object_sha256": "afc0c7db304251511eef5e223f027dc6314dd70895955ff1a1cb0a461dbfdbe9", + "program_id": "mfp3554", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "00b2ffe1073feede8447bac439ac7842c1ba0841cb6ee38ec1b5e06c3a3dc22e", + "program_context_pointers": [ + "/entries/46/papers/153", + "/entries/46/subsessions/9/items/21" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835591", + "title": "DP\n 2\n -NeRF: Dual-Prior Probabilistic Neural Radiance Fields for Event-Guided Motion Deblurring", + "authors": [ + "Jiafeng Huang", + "Wenzhen Jia", + "Shilong Dong", + "Shengjie Zhao" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-007.json", + "source_position": 97, + "registry_object_sha256": "145025e784ec251247f8cd86d0dbee1749666ec71f35566c20e41edd0a2347ce", + "program_id": "mfp3783", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "0b2d267736cfe506de666618820e2934c23395bd25dab63c5a347710445a361c", + "program_context_pointers": [ + "/entries/33/papers/158", + "/entries/33/subsessions/10/items/2" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835667", + "title": "MMCoT\n 3\n : Empowering 3D Scene Understanding with 3D-Text Interleaved Multi-Modal Chain-of-Thought Reasoning", + "authors": [ + "Yongqi Shan", + "Yifan Wang", + "Yunzhi Zhuge", + "Lijun Wang", + "Huchuan Lu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-005.json", + "source_position": 209, + "registry_object_sha256": "8b5ad8d891692e441b6e210304dcd35103835fd35d086c860c7611bf5385a6d9", + "program_id": "mfp4173", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "4924721a2c3328e6bca661654569bca86d3fa69aa5f6c9e3719f827c5f0d7639", + "program_context_pointers": [ + "/entries/88/papers/114", + "/entries/88/subsessions/8/items/6" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835763", + "title": "AudioKV:\n KV Cache Eviction in Efficient Large Audio Language Models", + "authors": [ + "Yuxuan Wang", + "He Peize", + "Xiyan Gui", + "Xiaoqian Liu", + "Junhao He", + "Xuyang Liu", + "Xuming Hu", + "Linfeng Zhang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-006.json", + "source_position": 192, + "registry_object_sha256": "70054b0a5a2586fd08e75494eee08a26498730554601a5fe2e4efddcaad309de", + "program_id": "mfp4778", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "274b360da553dfc7afdb9072505c2957cec4213e096373e2edbf58236a3fc3a2", + "program_context_pointers": [ + "/entries/129/papers/100", + "/entries/129/subsessions/7/items/16" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835895", + "title": "F\n 2\n -Adapter: A Focus-and-Forget Framework for Generalizable CLIP-based Face Forgery Detection", + "authors": [ + "Zan Gao", + "Xin Duan", + "Nian Hu", + "Hao Jia", + "Yibo Zhao", + "Cong Liu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-001.json", + "source_position": 39, + "registry_object_sha256": "16bdbcc4398df4d9556c14d4d4203127349070669f5d0e1c4bf87756e241b698", + "program_id": "mfp5537", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "b3ef626b66d93ac37a73caf15a4cf9ff92241209c80ee66be5575e6b56f9c899", + "program_context_pointers": [ + "/entries/88/papers/184", + "/entries/88/subsessions/11/items/6" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3835959", + "title": "UPR-VAD:\n U\n ncertainty-guided Signal\n P\n urification and\n R\n egularization for Weakly-Supervised\n V\n ideo\n A\n nomaly\n D\n etection", + "authors": [ + "Weiliang Huang", + "Weichao Cai", + "Bob Zhang", + "Chao Huang", + "Biao Xue", + "Rongxin Zhang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-007.json", + "source_position": 21, + "registry_object_sha256": "e96f9e9d67cce31a9235bdc7dc5dba551676f3d36ba72654ae3d5a98ae8d5174", + "program_id": "mfp5823", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "d8469bcb68f8cc19070f3bf9a866c3377ba27fe6445b5bf9337b7599a6881592", + "program_context_pointers": [ + "/entries/88/papers/66", + "/entries/88/subsessions/5/items/6" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3836215", + "title": "BLM\n 1\n : A Boundless Large Model for Cross-Space, Cross-Task, and Cross-Embodiment Learning", + "authors": [ + "Wentao Tan", + "Bowen Wang", + "Heng Zhi", + "Chenyv Liu", + "Zhe Li", + "Jian Liu", + "Zengrong Lin", + "Yukun Dai", + "Yipeng Chen", + "Wenjie Yang", + "Enci Xie", + "Hao Xue", + "Baixu Ji", + "Chen Xu", + "Zhibin Wang", + "Tianshi Wang", + "Lei Zhu", + "Heng Tao Shen" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-003.json", + "source_position": 46, + "registry_object_sha256": "b4b51e5d0a4c0a330d01e7638c0bcc73871d750312081bafbd3b8a10529153f2", + "program_id": "mfp7332", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "aa458905209ad562117d72dd33d4d61e000b0d9efaa55dd9672919f40e6cc2fd", + "program_context_pointers": [ + "/entries/33/papers/13", + "/entries/33/subsessions/1/items/1", + "/entries/47/papers/0" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3836272", + "title": "SEE: Structure-aware Exploring & Exploiting for Long-horizon GUI Agent Trajectory Synthesis", + "authors": [ + "Zhuohang Fan", + "Beichen Zhang", + "Yuanfa Li", + "Changqiao Wu", + "Wei Liu", + "Jian Luan", + "Weigang Zhang" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-002.json", + "source_position": 127, + "registry_object_sha256": "66eebbdef9e2d07fe08f3cb41cd577095f282f1256bdb73fe1e0d4c5d6e55d51", + "program_id": "mfp7635", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "df11d727337f397bab72977bcdce4246b388ef0c8e1198ac47fdc32e68a3aff4", + "program_context_pointers": [ + "/entries/88/papers/131", + "/entries/88/subsessions/8/items/23" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3836293", + "title": "Video\n *\n Bench: Benchmarking Fine-grained Video Perception with \"Think with Videos\"", + "authors": [ + "Ruizhe Chen", + "Zhiting Fan", + "Yang Shi", + "Enxin Song", + "Wenhao Chai", + "Tongkun Guan", + "Songtao Jiang", + "Ruilin Luo", + "Yuanxing Zhang", + "Shuai Bai", + "Sibo Song", + "Zhibo Yang", + "Zuozhu Liu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-002.json", + "source_position": 148, + "registry_object_sha256": "7332426edb3d15499f175fb98c093b78df21ea95fdd9cfea33da5e925cbec171", + "program_id": "mfp7732", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "f6e4df110fe85d7f6d34eb712cd338443c2e34e924393a4e65e07a0c7b152c9e", + "program_context_pointers": [ + "/entries/46/papers/198", + "/entries/46/subsessions/11/items/20" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3836447", + "title": "Insert In Style\n : A Zero-Shot Generative Framework for Harmonious Cross-Domain Object Composition", + "authors": [ + "Raghu Vamsi Chittersu", + "Yuvraj Singh Rathore", + "Pranav Adlinge", + "Eswar Sai Pashaula", + "Kunal Swami" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-002.json", + "source_position": 183, + "registry_object_sha256": "92b2c094b96cebbace0150805b65b0a1bc3ca99ed42124bcb8de442a81094e1b", + "program_id": "mfp8648", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "39d3f21f774010980f8b30f2fd800b5243ea7ddc6e091b4d8769de3dee622273", + "program_context_pointers": [ + "/entries/107/papers/4", + "/entries/117/papers/30", + "/entries/117/subsessions/2/items/6" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3836600", + "title": "EEGaze\n : Multimodal Attention Prediction in Visual Search via EEG-Based Saliency Refinement", + "authors": [ + "Mansi Sharma", + "Dinar Orazgaliyev", + "Antonio Krüger", + "Philipp Müller" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-003.json", + "source_position": 140, + "registry_object_sha256": "30b99b234aa2b944d6890119b2524d3b3c3c425bde22f20e813b9517b8f8536b", + "program_id": "mfp9815", + "track_literal": "Main Track", + "track": "main", + "abstract_sha256": "47dcb456aee600663d0df59429c1745b335dbae9bdef2dd1063265429fff49fb", + "program_context_pointers": [ + "/entries/46/papers/219", + "/entries/46/subsessions/12/items/17" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3838656", + "title": "MedMKG:\n Benchmarking Medical Knowledge Exploitation with Multimodal Knowledge Graph", + "authors": [ + "Xiaochen Wang", + "Yuan Zhong", + "Lingwei Zhang", + "Lisong Dai", + "Ting Wang", + "Fenglong Ma" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-003.json", + "source_position": 225, + "registry_object_sha256": "b910b66541ca65127646d35d0cd5737a56ae2e0f311a79526005dec75fca739f", + "program_id": "mds177", + "track_literal": "Datasets", + "track": "datasets", + "abstract_sha256": "5bee3b5e9ff60690509d1f34953f3dcc185c50861e112d2c4a3baf5673f6b945", + "program_context_pointers": [ + "/entries/46/papers/23", + "/entries/46/subsessions/1/items/11" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3839012", + "title": "EULALens\n : SEE What YOU Agree!", + "authors": [ + "Abhijeet Narang", + "Hagar Elbatanouny", + "Vanishri Shekhawat", + "Hasan Al-Nashash", + "Abhinav Dhall", + "Usman Tariq" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-007.json", + "source_position": 238, + "registry_object_sha256": "81dd29dafd6228f51ae2f32e2bb713272a88df78c78347b9170b54f83db00c84", + "program_id": "mde41", + "track_literal": "Demo and Video", + "track": "demo_and_video", + "abstract_sha256": "9acbb264b8019916aad0845aeae4eb5a9c94903ac5bca3cfb612c857306f6d55", + "program_context_pointers": [ + "/entries/94/papers/7" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3839024", + "title": "ThreatVerse:\n A Closed-Loop World Model Simulator for Autonomous Driving Safety Evaluation", + "authors": [ + "Jiangfan Liu", + "Zexuan Cui", + "Tianyuan Zhang", + "Zonglei Jing", + "Yaoyuan Zhang", + "Jiakai Wang", + "Jinyang Guo", + "Yuqing Ma", + "Lu Sheng", + "Aishan Liu", + "Xianglong Liu" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-008.json", + "source_position": 0, + "registry_object_sha256": "d0571f4562b30e4e7fb2e3aa826068ac14a540d1ae70e8cf0b11fd73c67f26ee", + "program_id": "mde15", + "track_literal": "Demo and Video", + "track": "demo_and_video", + "abstract_sha256": "02fc9b31d2b7667ea13f00c74f923a375373042171aaa1ff1f2010ee9697211a", + "program_context_pointers": [ + "/entries/49/papers/4" + ], + "single_name_positions": [] + }, + { + "doi": "10.1145/3767308.3839028", + "title": "FinAlogy:\n A Visual Analogy Retrieval System for Financial K-Line Analysis", + "authors": [ + "Xiaoyan Yu", + "Ziya Zhao", + "Yunshan Ma", + "Yifan Wei", + "Dianbo Sui", + "Tat-Seng Chua" + ], + "kind": "unique_program_format_title_fullauthor_agreement_not_identity_approval", + "source_page": "page-008.json", + "source_position": 4, + "registry_object_sha256": "b18c8e2821873a59318f6f09d1f453dd02195ccedad5c8e0ce1da60ac07466fb", + "program_id": "mde02", + "track_literal": "Demo and Video", + "track": "demo_and_video", + "abstract_sha256": "22f978dd02c05439663c5724f25bb997098fd85f5558d01d836711252d119931", + "program_context_pointers": [ + "/entries/49/papers/0" + ], + "single_name_positions": [] + } + ], + "author_disagreement_hold": { + "doi": "10.1145/3767308.3836308", + "program_id": "mfp7801" + }, + "scope": "Onlytitlecomparisonchanges;originalregistryandprogramfixturesremainunchanged;sourceABandtrackcomefromtheexplicitprogramID.", + "held_gold": { + "doi": "10.1145/3767308.3836308", + "title": "Dynamic Late Fusion Multiple Kernel\n k\n -means Clustering based on Adaptive Sample Selection and Alignment", + "authors": [ + "Zhirong Tian", + "Yi Zhang", + "Weixuan Liang", + "Miaomiao Li", + "Xinwang Liu" + ], + "kind": "program_format_author_version_requires_review", + "source_page": "page-001.json", + "source_position": 82, + "registry_object_sha256": "26a3c1715d276b5fbe988596e3115091086158f6ef12600a36e5206785fc9dfb", + "program_id": null, + "track_literal": null, + "track": "proceedings_unclassified", + "abstract_sha256": null, + "program_context_pointers": [], + "single_name_positions": [] + } +} diff --git a/tests/test_mm_publications2026.py b/tests/test_mm_publications2026.py index 3614a7e..9dd46ed 100644 --- a/tests/test_mm_publications2026.py +++ b/tests/test_mm_publications2026.py @@ -19,6 +19,9 @@ class MMPublications2026Tests(unittest.TestCase): fixtures=Path(__file__).parent/'fixtures' cls.fixture=json.loads(gzip.decompress((fixtures/'mm2026-publications.json.gz').read_bytes())) cls.program_fixture=json.loads(gzip.decompress((fixtures/'mm2026-source.json.gz').read_bytes())) + cls.format_gold=json.loads((fixtures/'mm2026-title-formatting-review.json').read_text()) + overrides={g['doi']:g for g in [*cls.format_gold['gold'],cls.format_gold['held_gold']]} + cls.gold=[overrides.get(g['doi'],g) for g in cls.fixture['independent_gold']] cls.temp=tempfile.TemporaryDirectory();cls.root=Path(cls.temp.name) cls.materialize(cls.root/'registry',cls.fixture['files']) cls.materialize(cls.root/'program',cls.program_fixture['files']) @@ -38,8 +41,8 @@ class MMPublications2026Tests(unittest.TestCase): def test_all1918_publications_against_independent_registry_program_gold(self): self.assertEqual(len(self.rows),1918);self.assertEqual(len(self.by_doi),1918) - self.assertEqual(set(self.by_doi),{g['doi'] for g in self.fixture['independent_gold']}) - for g in self.fixture['independent_gold']: + self.assertEqual(set(self.by_doi),{g['doi'] for g in self.gold}) + for g in self.gold: r=self.by_doi[g['doi']];m=r['metadata'];p=m['program_fields'] self.assertEqual((r['title'],r['authors'],r['track']),(g['title'],g['authors'],g['track'])) self.assertEqual(m['program_match_kind'],g['kind']) @@ -51,22 +54,22 @@ class MMPublications2026Tests(unittest.TestCase): def test_abstract_source_is_separate_and_verbatim_not_fabricated_registry_abstract(self): count=0 - for g in self.fixture['independent_gold']: + for g in self.gold: r=self.by_doi[g['doi']];m=r['metadata'] self.assertFalse(m['registry_abstract_available']);self.assertFalse(m['registry_fields'].get('abstract')) if g['abstract_sha256'] is not None: count+=1;self.assertEqual(hashlib.sha256(r['abstract'].encode()).hexdigest(),g['abstract_sha256']) self.assertEqual(r['abstract'],m['program_fields']['abstract']) - self.assertEqual(m['abstract_source'],'official_program_exact_metadata_correspondence') + self.assertEqual(m['abstract_source'],'official_program_reviewed_title_formatting_correspondence' if g['kind'].startswith('unique_program_format_') else 'official_program_exact_metadata_correspondence') else: self.assertIsNone(r['abstract']);self.assertEqual(m['abstract_source'],'not_available_from_reviewed_sources') - self.assertEqual(count,1717);self.assertEqual(self.report['output_missing_abstract_count'],201) + self.assertEqual(count,1743);self.assertEqual(self.report['output_missing_abstract_count'],175) def test_all_program_tracks_and_4326_contexts_survive_in_matches_or_unmatched(self): programs={r['metadata']['program_fields']['paper_id']:r['metadata']['program_fields'] for r in self.rows if r['metadata']['program_fields']} for p in self.report['unmatched_program']: self.assertNotIn(p['paper_id'],programs);programs[p['paper_id']]=p - self.assertEqual(len(programs),2143);self.assertEqual(self.report['unmatched_program_count'],417) + self.assertEqual(len(programs),2143);self.assertEqual(self.report['unmatched_program_count'],391) gold=self.program_fixture['independent_inventory']['inventory'] self.assertEqual(set(programs),{g['paper_id'] for g in gold}) for g in gold: @@ -91,7 +94,7 @@ class MMPublications2026Tests(unittest.TestCase): for track in ('interactive_art','industry_talk','demo_and_video'): self.assertTrue(any(r['track']==track for r in self.rows)) variants=[r for r in self.rows if r['track']=='proceedings_unclassified'] - self.assertEqual(len(variants),192) + self.assertEqual(len(variants),166) self.assertTrue(all(r['metadata']['program_fields'] is None and r['abstract'] is None for r in variants)) def test_original_bibliography_evidence_kept_without_duplicating_reference_payload(self): @@ -113,7 +116,7 @@ class MMPublications2026Tests(unittest.TestCase): def test_eight_single_name_fields_preserved_without_splitting(self): n=0 - for g in self.fixture['independent_gold']: + for g in self.gold: r=self.by_doi[g['doi']];flags=r['metadata']['review_flags'] actual=[x['author_position'] for x in flags if x['kind']=='single_registered_name_field_not_split'] self.assertEqual(actual,g['single_name_positions']);n+=len(actual) @@ -123,13 +126,62 @@ class MMPublications2026Tests(unittest.TestCase): def test_structured_title_markup_retained_and_flagged_without_flattening(self): count=0 - for g in self.fixture['independent_gold']: + for g in self.gold: if re.search(r'<[^>]+>|&[A-Za-z#0-9]+;',g['title']): count+=1;r=self.by_doi[g['doi']] self.assertEqual(r['title'],g['title']) self.assertIn({'kind':'literal_title_markup_or_entity_retained_requires_review'},r['metadata']['review_flags']) self.assertGreater(count,0);self.assertEqual(count,self.report['registered_title_markup_flags']) + + def test_formatting_gold_preserves_raw_fields_and_copies_exact_program_evidence(self): + self.assertEqual(len(self.format_gold['gold']),26) + for gold in self.format_gold['gold']: + row=self.by_doi[gold['doi']];m=row['metadata'];p=m['program_fields'] + self.assertEqual((row['title'],row['authors']),(gold['title'],gold['authors'])) + self.assertEqual((p['paper_id'],row['track']),(gold['program_id'],gold['track'])) + self.assertEqual(hashlib.sha256(row['abstract'].encode()).hexdigest(),gold['abstract_sha256']) + comparison=m['program_title_formatting_comparison'] + self.assertTrue(comparison['raw_titles_preserved']);self.assertFalse(comparison['work_identity_approved']) + self.assertEqual(set(comparison['program_comparison_texts']),{p['paper_id']}) + self.assertFalse(m['scientific_work_identity_approved']) + held=self.by_doi[self.format_gold['author_disagreement_hold']['doi']] + self.assertEqual(held['metadata']['program_title_candidates'],['mfp7801']) + self.assertEqual(held['metadata']['program_exact_candidates'],[]) + self.assertEqual(held['metadata']['program_match_kind'],'program_format_author_version_requires_review') + self.assertIsNone(held['abstract']);self.assertEqual(held['track'],'proceedings_unclassified') + + def test_reviewed_formatting_keeps_superscript_subscript_and_decodes_entities_once(self): + self.assertEqual(c.formatting_text('A H2 B1 & C'),'A H2 B1 & C') + self.assertEqual(c.formatting_text('A &amp; B'),'A & B') + + def test_unknown_attributes_unbalanced_or_encoded_markup_not_discarded(self): + for title in ['A B','A B','A B','A B', + 'A B','A B','A','A <i>B</i>', + 'A B','']: + with self.subTest(title=title):self.assertIsNone(c.formatting_text(title)) + + def test_ambiguous_program_and_author_multiset_differences_remain_held(self): + def observation(pid,authors): + return {'paper_id':pid,'title':'M3 Model','authors':authors,'track':'Main Track'} + rows=[{'title':'M3 Model','authors':['A','B']}] + kinds,_=c.match_program(rows,[observation('p1',['A','B']),observation('p2',['B','A'])]) + self.assertEqual(kinds,{'multiple_program_ids_same_title_authors_hold':1}) + self.assertNotIn('program',rows[0]) + rows=[{'title':'M3 Model','authors':['A','A','B']}] + c.match_program(rows,[observation('p1',['A','B'])]) + self.assertEqual(rows[0]['kind'],'program_format_author_version_requires_review') + self.assertNotIn('program',rows[0]) + + def test_literal_title_author_disagreement_does_not_use_formatting_bypass(self): + rows=[{'title':'M3 Model','authors':['A','B']}] + observations=[{'paper_id':'literal','title':'M3 Model','authors':['Other'],'track':'Main Track'}, + {'paper_id':'formatted','title':'M3 Model','authors':['A','B'],'track':'Main Track'}] + c.match_program(rows,observations) + self.assertEqual(rows[0]['kind'],'program_author_version_requires_review') + self.assertEqual(rows[0]['program_title_candidates'],['literal']);self.assertNotIn('program',rows[0]) + + def test_corrupt_registry_parent_or_program_keeps_failure_not_complete_output(self): for kind in ('registry_manifest','registry_page','parent','program'): with tempfile.TemporaryDirectory() as tmp: @@ -150,9 +202,12 @@ class MMPublications2026Tests(unittest.TestCase): c.collect(self.out,registry_snapshots=self.root/'registry',program_snapshots=self.root/'program',parent_snapshot=self.root/'registry/parent.json') def test_complete_report_retains_source_partitions_not_import_or_final_dedup_claim(self): - self.assertEqual(self.report['program_match_counts'],{'unique_program_title_fullauthor_agreement_not_identity_approval':1726,'program_author_version_requires_review':122,'no_exact_program_title_requires_review':70}) + self.assertEqual(self.report['program_match_counts'],{'unique_program_title_fullauthor_agreement_not_identity_approval':1726,'unique_program_format_title_fullauthor_agreement_not_identity_approval':26,'program_author_version_requires_review':122,'program_format_author_version_requires_review':1,'no_exact_program_title_requires_review':43}) self.assertEqual(self.report['registry_missing_abstract_count'],1918) - self.assertEqual(self.report['abstracts_from_exact_program_correspondence'],1717) + self.assertEqual(self.report['abstracts_from_exact_program_correspondence'],1743) + self.assertEqual(self.report['abstracts_from_literal_program_correspondence'],1717) + self.assertEqual(self.report['abstracts_from_reviewed_title_formatting'],26) + self.assertEqual(self.report['unmatched_program_count'],391) self.assertEqual(sum(f['count'] for f in self.report['files']),1918) self.assertEqual(self.report['approved_for_import'],0);self.assertFalse(self.report['database_written']) self.assertFalse(self.report['pdfs_downloaded']);self.assertEqual(self.report['network_requests'],0) @@ -196,7 +251,7 @@ class MMPublications2026Tests(unittest.TestCase): meta=json.loads(got[5]);self.assertEqual(meta.pop('record_identifiers'),{'doi':doi,c.SOURCE:doi});self.assertEqual(meta,row['metadata']) tags=conn.execute("SELECT t.namespace,t.value,count(*) FROM paper_tags pt JOIN tags t ON t.id=pt.tag_id GROUP BY t.namespace,t.value").fetchall() self.assertEqual(sum(r[2] for r in tags if r[0]==c.SOURCE+':crossref:type'),1918) - self.assertEqual(sum(r[2] for r in tags if r[0]==c.SOURCE+':mm_program:track'),1726) + self.assertEqual(sum(r[2] for r in tags if r[0]==c.SOURCE+':mm_program:track'),1752) self.assertEqual({r[0] for r in tags},{c.SOURCE+':crossref:type',c.SOURCE+':mm_program:track'}) decisions=conn.execute("SELECT d.status,d.decision_date,count(*) FROM decisions d JOIN submissions s ON s.id=d.submission_id JOIN source_records r ON r.id=s.source_record_id JOIN sources src ON src.id=r.source_id WHERE src.name=? GROUP BY d.status,d.decision_date",(c.SOURCE,)).fetchall() self.assertEqual([tuple(r) for r in decisions],[('accepted',None,1918)])