{"id":"W7092317075","doi":"","title":"FACTS: Table Summarization via Offline Template Generation with Agentic Workflows","year":2025,"lang":"","type":"article","venue":"arXiv (Cornell University)","topic":"Data Quality and Management","field":"Decision Sciences","cited_by":0,"is_retracted":false,"has_abstract":true,"ca_institutions":"Mila - Quebec Artificial Intelligence Institute","funders":"","keywords":"Automatic summarization; Table (database); Workflow; Executable; Natural language; Key (lock); Robustness (evolution)","routes":{"ca_aff":true,"ca_fund":false,"ca_venue":false,"about_ca":false,"invisible_to_affiliation_only":false},"retraction":null,"screen":null,"direct_labels":[],"prediction":{"model_version":"codex-gemma-dda1882f352a","candidate_categories":["metaepi_narrow","insufficient_payload"],"consensus_categories":["insufficient_payload"],"category_scores_codex":[0.001698918,0.0003883153,0.0004653861,0.0007957472,0.0008312522,0.0006906922,0.001244531,0.0002042098,0.001735035],"category_scores_gemma":[0.0002312163,0.0003792058,0.0001302969,0.005338068,0.0002798779,0.002155202,0.0006614904,0.0002540306,0.0008481047],"about_ca_system_candidate":false,"about_ca_system_consensus":false,"about_ca_system_score_codex":0.0002580811,"about_ca_system_score_gemma":0.0002809272,"about_ca_topic_candidate":false,"about_ca_topic_consensus":false,"about_ca_topic_score_codex":0.0004446479,"about_ca_topic_score_gemma":0.001219451,"domain_scores_codex":[0.9963262,0.0005147818,0.0006910149,0.001517549,0.0003872921,0.0005631782],"domain_scores_gemma":[0.9970916,0.0002900919,0.0004727032,0.001443835,0.0004944806,0.0002072927],"domain_codex":null,"domain_gemma":null,"domain_candidate":null,"domain_consensus":null,"study_design_codex":"simulation_or_modeling","study_design_gemma":"simulation_or_modeling","study_design_scores_codex":[0.0009111525,0.001078052,0.01368392,0.000199597,0.0007749544,0.0004383631,0.0005117977,0.6514819,0.0008799168,0.2652919,0.03521484,0.02953358],"study_design_scores_gemma":[0.002477529,0.0002499529,0.001616279,0.0002135016,0.0006503699,0.000002726217,0.001163615,0.8760082,0.0007587245,0.01390688,0.1022245,0.0007277353],"study_design_candidate":"simulation_or_modeling","study_design_consensus":"simulation_or_modeling","genre_codex":"methods","genre_gemma":"empirical","genre_scores_codex":[0.1080773,0.0001014496,0.8742929,0.0004166556,0.0009689983,0.0006717091,0.00007393734,0.00006393819,0.01533318],"genre_scores_gemma":[0.8927749,0.000368962,0.0005037714,0.0006143355,0.00007135668,0.000001162442,0.0002064903,0.00001509501,0.1054439],"genre_candidate":"empirical","genre_consensus":null,"teacher_disagreement_score":0.8737891,"threshold_uncertainty_score":0.9999298,"prediction_status":"machine_predicted_unvalidated"},"machine_scores":{"provisional":true,"baseline":true,"maturity_gate_passed":false,"score_opus":0.1845085958701661,"score_gpt":0.2489525282214207,"score_spread":0.06444393235125456,"validation_status":"score_only:v0-immature-baseline","note":"Baseline scores from an immature model (maturity gate not passed). Scores rank; they never assert a category."}}