{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2608.17550", "kind": "paper", "url": "https://arxiv.org/abs/2608.17550"}, {"id": "gh:ariking11/cadp-bench", "kind": "repo", "url": "https://github.com/AriKing11/CADP-Bench"}], "authors": ["Rihui Jin", "Jun Wang", "chengyuan zhu", "Liang Mingyu", "Yue Gao", "Li Yunxuan", "Kuicai Dong", "Guilin Qi", "Lin Ren", "Yongrui Chen", "Xinbang Dai", "Jiaqi Li", "Tongtong Wu", "Gholamreza Haffari"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.17550", "source_url": "https://arxiv.org/abs/2608.17550"}, "categories": ["Academic document parsing", "Chart-to-code generation", "Multimodal", "Multimodal code generation", "Scientific Research & AI for Science", "Structured document reconstruction", "cs.CV"], "description": {"en": "CADP-Bench evaluates multimodal LLMs on compilable academic document parsing. It includes expert-verified full academic pages with tightly coupled text and structured elements, assessed through a re-injection compilation protocol."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.17550", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.17550", "title": "CADP-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Southeast University", "Nanyang Technological University", "Nanjing University", "Monash University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2608.17550", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2608.17550", "source_url": "https://openalex.org/W7203754913"}, "key": "claire-radar:2608.17550", "modality": null, "name": "CADP-Bench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_cadp-bench_c4c34058", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "6582d0b7efec8ac174cb960173d65a8003462819bb62389338f7947fd77e8028", "relation": "introduces", "review_state": "source-reviewed", "reviewed_at": "2026-08-20", "source_url": "https://arxiv.org/abs/2608.17550"}, "publisher": null, "released": "2026-08-18", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.17550", "source_url": "https://arxiv.org/abs/2608.17550"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-17550", "source": "claire_radar", "source_benchmark_id": "2608.17550", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["Scientific Research & AI for Science"], "area": "Multimodal", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/AriKing11/CADP-Bench", "githubScope": "benchmark_repo", "githubStars": 1, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": null, "hfPaperUrl": "https://huggingface.co/papers/2608.17550", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "state": "unavailable"}}, "status": "fresh"}, "availability": {"evaluatorStatus": "available", "githubStatus": "available", "hfDatasetStatus": "sample_only", "paperStatus": "available", "submissionStatus": "not_found"}, "capabilities": ["Academic document parsing", "Structured document reconstruction", "Chart-to-code generation", "Multimodal code generation"], "confidence": "High", "construction": "Unknown", "constructionDetail": "CADP-Bench asks models to reconstruct complete academic pages as compilable LaTeX and executable Python, then scores both structure and rendered output.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "85b405ab2d375c2799108825ab66f088802f9ba6930377405a31d41e0d220d9e", "model": "deepseek-v4-flash"}, "curation": {"reviewedAt": "2026-08-20", "sources": ["https://arxiv.org/abs/2608.17550", "https://arxiv.org/html/2608.17550v1", "https://github.com/AriKing11/CADP-Bench"], "state": "source-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "CADP-Bench evaluates multimodal LLMs on compilable academic document parsing. It includes expert-verified full academic pages with tightly coupled text and structured elements, assessed through a re-injection compilation protocol.", "detail": {"protocol": {"primaryMetric": "Pixel Similarity with structural, reading-order and execution metrics", "tasks": "1,630 full-page samples", "version": "v1"}, "taskBreakdown": ["Computer Science", "Physics", "Economics", "Quantitative Biology", "Statistics"]}, "domainCuration": {"method": "primary-source-review-v1", "state": "reviewed"}, "evaluationMode": "score_submission", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "To support this setting, we introduce CADP-Bench, an expert-verified benchmark of full academic pages containing tightly coupled text and multiple SAE types, evaluated through a re-injection compilation protocol."}, "familyId": "bmf_a82a8535ed6e", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/AriKing11/CADP-Bench", "url": "https://github.com/AriKing11/CADP-Bench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/AriKing11/CADP-Bench", "url": "https://huggingface.co/papers/2608.17550"}, {"basis": "record-resource-link", "linkedUrl": "https://github.com/AriKing11/CADP-Bench", "url": "https://github.com/AriKing11/CADP-Bench"}, {"basis": "record-resource-link", "linkedUrl": "https://github.com/AriKing11/CADP-Bench/tree/main/data", "url": "https://github.com/AriKing11/CADP-Bench/tree/main/data"}], "scope": "benchmark_repo", "url": "https://github.com/AriKing11/CADP-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_cadp-bench_c4c34058", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/AriKing11/CADP-Bench", "data": "https://github.com/AriKing11/CADP-Bench/tree/main/data", "hfPaper": "https://huggingface.co/papers/2608.17550", "paper": "https://arxiv.org/abs/2608.17550", "pdf": "https://arxiv.org/pdf/2608.17550", "project": "https://github.com/AriKing11/CADP-Bench", "report": "https://arxiv.org/abs/2608.17550"}, "metrics": [], "motivation": "Academic papers are a primary carrier of scientific knowledge, yet most of this knowledge remains locked in PDFs that are optimized for human reading rather than machine use.", "name": "CADP-Bench", "oneLine": "Evaluates MLLMs that reconstruct full academic pages as compilable LaTeX and executable Python from page images.", "paperTitle": "Code as Representation: A Compilable Parsing Paradigm for Academic Documents", "primaryDomain": "Scientific Research & AI for Science", "publication": {"evidence": "Accepted by ACM MM 2026", "evidenceLevel": "author-claim", "evidenceUrl": "https://arxiv.org/abs/2608.17550", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "ACM MM 2026", "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {"90d": {"confidence": "Low", "coverage": 0.55, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.09252268850709358, "value": 1}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": null, "value": null}}, "confidence": "Low", "coverage": 0.55, "rank": 336, "score": 28}, "method": "attention-ranking-v13", "rank": 336, "score": 28}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-08-18", "source": {"authors": ["Rihui Jin", "Jun Wang", "chengyuan zhu", "Liang Mingyu", "Yue Gao", "Li Yunxuan", "Kuicai Dong", "Guilin Qi", "Lin Ren", "Yongrui Chen", "Xinbang Dai", "Jiaqi Li", "Tongtong Wu", "Gholamreza Haffari"], "categories": ["cs.CV", "cs.CL"], "id": "2608.17550", "title": "Code as Representation: A Compilable Parsing Paradigm for Academic Documents", "type": "arxiv", "url": "https://arxiv.org/abs/2608.17550"}, "sourceUpdatedAt": "2026-08-18T09:10:43Z", "topics": ["cs.CV"], "venueAttempts": [{"decisionRaw": "Accepted by ACM MM 2026", "evidence": [{"level": "author-claim", "observedAt": "2026-10-14T01:46:58.058199Z", "rawValue": "Accepted by ACM MM 2026", "sourceType": "arxiv-comments", "sourceUrl": "https://arxiv.org/abs/2608.17550"}], "reviewStatus": "accepted", "venueName": "ACM MM 2026"}], "whyItMatters": "Provides a structured evaluation for structure-aware scientific document parsing, measuring the fidelity of executable reconstructions, which is crucial for machine-readable scientific knowledge."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
