{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2606.09323", "kind": "paper", "url": "https://arxiv.org/abs/2606.09323"}, {"id": "gh:logo-cuhksz/trl-bench", "kind": "repo", "url": "https://github.com/LOGO-CUHKSZ/TRL-Bench"}], "authors": ["Wei Pang", "Xiangru Jian", "Hehan Li", "Zhixuan Yu", "Alex Xue", "Jinyang Li", "Zhengyuan Dong", "Xinjian Zhao", "Hao Xu", "Chao Zhang", "Reynold Cheng", "M. Tamer Özsu", "Tianshu Yu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2606.09323", "source_url": "https://arxiv.org/abs/2606.09323"}, "categories": ["General AI", "Language & Knowledge", "cs.AI"], "description": {"en": "TRL-Bench evaluates tabular encoders at the representation level by exporting row-, column-, or table-level embeddings through each encoder's supported wrapper and probing them with shared lightweight heads across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment) covering 16 tasks and 20 models."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2606.09323", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2606.09323", "title": "TRL-Bench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["Chinese University of Hong Kong, Shenzhen", "University of Waterloo", "University of Hong Kong", "The University of Sydney"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2606.09323", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2606.09323", "source_url": "https://openalex.org/W7164030619"}, "key": "claire-radar:2606.09323", "modality": null, "name": "TRL-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_trl-bench_b244de7b", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "e53fb4efa6323d74fd23df3fe3580dd3d914c5f7f87f4be21749f602fd91075c", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2606.09323"}, "publisher": null, "released": "2026-06-08", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2606.09323", "source_url": "https://arxiv.org/abs/2606.09323"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2606-09323", "source": "claire_radar", "source_benchmark_id": "2606.09323", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/LOGO-CUHKSZ/TRL-Bench", "githubScope": "benchmark_repo", "githubStars": 10, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 54, "hfPaperUrl": "https://huggingface.co/papers/2606.09323", "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", "lastSuccessfulDate": "2026-09-24", "state": "stale"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "75d17630044b0d81b27f75e8919bf9f8009ba719fd915f9545b8ae5a650f9be4", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "TRL-Bench evaluates tabular encoders at the representation level by exporting row-, column-, or table-level embeddings through each encoder's supported wrapper and probing them with shared lightweight heads across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment) covering 16 tasks and 20 models.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "We introduce TRL-Bench, a multi-granular tabular representation learning (TRL) benchmark that standardizes cross-paradigm representation-level evaluation: each encoder exports row-, column-, or table embeddings through its supported wrapper, and shared lightweight heads probe them across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment spanning all three granularities)."}, "familyId": "bmf_722ae64b1c7b", "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/LOGO-CUHKSZ/TRL-Bench", "url": "https://github.com/LOGO-CUHKSZ/TRL-Bench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/LOGO-CUHKSZ/TRL-Bench", "url": "https://huggingface.co/papers/2606.09323"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/LOGO-CUHKSZ/TRL-Bench", "url": "https://arxiv.org/abs/2606.09323"}], "scope": "benchmark_repo", "url": "https://github.com/LOGO-CUHKSZ/TRL-Bench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_trl-bench_b244de7b", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/LOGO-CUHKSZ/TRL-Bench", "data": null, "hfPaper": "https://huggingface.co/papers/2606.09323", "paper": "https://arxiv.org/abs/2606.09323", "pdf": "https://arxiv.org/pdf/2606.09323", "project": null, "report": "https://arxiv.org/abs/2606.09323"}, "metrics": [], "motivation": "Tabular encoders are usually evaluated inside task-specific end-to-end pipelines, so models from different training paradigms are difficult to compare directly even when they operate on similar tabular signals.", "name": "TRL-Bench", "oneLine": "TRL-Bench evaluates tabular encoders at the representation level by exporting row-, column-, or table-level embeddings through each encoder's supported wrapper and probing them with shared lightweight heads across three suites: TRL-CTbench (column/table), TRL-Rbench (row), and TRL-DLTE (compositional Data-Lake Table Enrichment) covering 16 tasks and 20 models.", "paperTitle": "TRL-Bench: Standardizing Cross-Paradigm Representation-Level Evaluation of Tabular Encoders", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2606.09323", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "LOGO Lab, CUHK-Shenzhen", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/LOGO-CUHKSZ/TRL-Bench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-06-08", "source": {"authors": ["Wei Pang", "Xiangru Jian", "Hehan Li", "Zhixuan Yu", "Alex Xue", "Jinyang Li", "Zhengyuan Dong", "Xinjian Zhao", "Hao Xu", "Chao Zhang", "Reynold Cheng", "M. Tamer \\\"Ozsu", "Tianshu Yu"], "categories": ["cs.AI", "cs.DB"], "id": "2606.09323", "title": "TRL-Bench: Standardizing Cross-Paradigm Representation-Level Evaluation of Tabular Encoders", "type": "arxiv", "url": "https://arxiv.org/abs/2606.09323"}, "sourceUpdatedAt": "2026-06-08T10:39:25Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "Traditional end-to-end pipelines obscure the comparative quality of tabular encoders from different training paradigms. TRL-Bench provides a standardized protocol to isolate representation-level capability, enabling model selection based on task-specific strengths rather than a single aggregate score."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
