{"url":"/sota/visual-question-answering-on-gqa-test2019","task":{"name":"Visual Question Answering (VQA)","url":"/task/visual-question-answering","note":null},"dataset":{"name":"GQA Test2019","url":"/dataset/gqa"},"category":"Computer Vision","categories":["Computer Vision","Natural Language Processing"],"category_note":null,"description":"**Visual Question Answering (VQA)** is a task in computer vision that involves answering questions about an image. The goal of VQA is to teach machines to understand the content of an image and answer questions about it in natural language.\r\n\r\nImage Source: [visualqa.org](https://visualqa.org/)","description_from":"task","source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28","rank":"the archive's row order at snapshot; not re-ranked","rows_end_at":"2025-07-28","rows_withheld_as_spam":0,"metric_values":"the archive's strings, untouched"},"metrics":["Accuracy","Binary","Open","Consistency","Plausibility","Validity","Distribution"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"Accuracy":"higher","Binary":null,"Open":null,"Consistency":null,"Plausibility":null,"Validity":null,"Distribution":null}},"counts":{"rows":127,"rows_with_code":5,"rows_with_paper_page":6,"rows_dated":6,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"human","metrics":{"Accuracy":"89.3","Binary":"91.2","Consistency":"98.4","Distribution":"0.0","Open":"87.4","Plausibility":"97.2","Validity":"98.9"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"DREAM+Unicoder-VL (MSRA)","metrics":{"Accuracy":"76.04","Binary":"84.46","Consistency":"91.47","Distribution":"3.68","Open":"68.6","Plausibility":"83.75","Validity":"96.42"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"TRRNet (Ensemble)","metrics":{"Accuracy":"74.03","Binary":"82.12","Consistency":"89.0","Distribution":"1.29","Open":"66.89","Plausibility":"83.58","Validity":"96.76"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"MIL-nbgao","metrics":{"Accuracy":"73.81","Binary":"80.8","Consistency":"91.76","Distribution":"1.7","Open":"67.64","Plausibility":"83.9","Validity":"96.73"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"Kakao Brain","metrics":{"Accuracy":"73.33","Binary":"79.68","Consistency":"77.02","Distribution":"2.46","Open":"67.73","Plausibility":"83.7","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":6,"model":"Coarse-to-Fine Reasoning, Single Model","metrics":{"Accuracy":"72.14","Binary":"81.16","Consistency":"90.96","Distribution":"2.39","Open":"64.19","Plausibility":"84.81","Validity":"96.77"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"270","metrics":{"Accuracy":"70.23","Binary":"77.5","Consistency":"86.94","Distribution":"1.49","Open":"63.82","Plausibility":"83.77","Validity":"96.65"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":8,"model":"NSM ensemble (updated)","metrics":{"Accuracy":"67.55","Binary":"80.45","Consistency":"93.83","Distribution":"2.78","Open":"56.16","Plausibility":"84.16","Validity":"96.53"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":9,"model":"VinVL-DPT","metrics":{"Accuracy":"64.92","Binary":"82.63","Consistency":"94.37","Distribution":"5.11","Open":"49.29","Plausibility":"84.91","Validity":"96.64"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":10,"model":"VinVL+L","metrics":{"Accuracy":"64.85","Binary":"82.59","Consistency":"94.0","Distribution":"4.59","Open":"49.19","Plausibility":"84.91","Validity":"96.62"},"uses_additional_data":false,"paper_date":"2023-02-22","paper":"/paper/vinvl-l-enriching-visual-representation-with","paper_url":"https://ceur-ws.org/Vol-3349/","paper_title":"VinVL+L: Enriching Visual Representation with Location Context in VQA","code":"https://github.com/vyskocj/VinVL-L","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"Single Model","metrics":{"Accuracy":"64.65","Binary":"82.63","Consistency":"94.35","Distribution":"4.72","Open":"48.77","Plausibility":"84.98","Validity":"96.62"},"uses_additional_data":false,"paper_date":"2021-01-02","paper":"/paper/vinvl-making-visual-representations-matter-in","paper_url":"https://arxiv.org/abs/2101.00529v2","paper_title":"VinVL: Revisiting Visual Representations in Vision-Language Models","code":"https://github.com/microsoft/Oscar","n_code_links":7,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":1}},{"rank_in_archive_order":12,"model":"Wayne","metrics":{"Accuracy":"63.94","Binary":"80.84","Consistency":"91.54","Distribution":"4.69","Open":"49.03","Plausibility":"84.74","Validity":"96.56"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":13,"model":"Single","metrics":{"Accuracy":"63.2","Binary":"77.91","Consistency":"89.84","Distribution":"5.25","Open":"50.22","Plausibility":"85.15","Validity":"96.47"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":14,"model":"NSM single (updated)","metrics":{"Accuracy":"63.17","Binary":"78.94","Consistency":"93.25","Distribution":"3.71","Open":"49.25","Plausibility":"84.28","Validity":"96.41"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":15,"model":"LXR955, Ensemble","metrics":{"Accuracy":"62.71","Binary":"79.79","Consistency":"93.1","Distribution":"6.42","Open":"47.64","Plausibility":"85.21","Validity":"96.36"},"uses_additional_data":false,"paper_date":"2019-08-20","paper":"/paper/lxmert-learning-cross-modality-encoder","paper_url":"https://arxiv.org/abs/1908.07490v3","paper_title":"LXMERT: Learning Cross-Modality Encoder Representations from Transformers","code":"https://github.com/huggingface/transformers","n_code_links":9,"syntology":{"n_ran":4,"n_unverified":11,"n_samples":15,"n_pointer_only_licence":3}},{"rank_in_archive_order":16,"model":"MDETR","metrics":{"Accuracy":"62.45","Binary":"80.91","Consistency":"93.95","Distribution":"5.36","Open":"46.15","Plausibility":"84.15","Validity":"96.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":17,"model":"1-gqa","metrics":{"Accuracy":"62.44","Binary":"80.28","Consistency":"94.36","Distribution":"5.33","Open":"46.69","Plausibility":"84.91","Validity":"96.46"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":18,"model":"UCM","metrics":{"Accuracy":"61.49","Binary":"78.4","Consistency":"88.68","Distribution":"5.7","Open":"46.56","Plausibility":"84.85","Validity":"96.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":19,"model":"GRN","metrics":{"Accuracy":"61.22","Binary":"78.69","Consistency":"90.31","Distribution":"6.77","Open":"45.81","Plausibility":"85.43","Validity":"96.36"},"uses_additional_data":false,"paper_date":"2019-07-23","paper":"/paper/graph-reasoning-networks-for-visual-question","paper_url":"https://arxiv.org/abs/1907.09815v2","paper_title":"Bilinear Graph Networks for Visual Question Answering","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":20,"model":"lxmert-adv-txt","metrics":{"Accuracy":"61.12","Binary":"78.07","Consistency":"91.13","Distribution":"5.55","Open":"46.16","Plausibility":"84.8","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":21,"model":"lxmert-adv-txt","metrics":{"Accuracy":"61.1","Binary":"77.99","Consistency":"91.08","Distribution":"5.52","Open":"46.19","Plausibility":"84.82","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":22,"model":"MSM@MSRA","metrics":{"Accuracy":"61.09","Binary":"77.84","Consistency":"88.92","Distribution":"5.68","Open":"46.3","Plausibility":"85.49","Validity":"96.43"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":23,"model":"mlmbert","metrics":{"Accuracy":"61.05","Binary":"78.02","Consistency":"89.77","Distribution":"5.24","Open":"46.06","Plausibility":"84.95","Validity":"96.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":24,"model":"fisher","metrics":{"Accuracy":"60.98","Binary":"77.32","Consistency":"90.77","Distribution":"5.36","Open":"46.55","Plausibility":"84.93","Validity":"96.38"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":25,"model":"ckpt 19 exp 90","metrics":{"Accuracy":"60.95","Binary":"78.41","Consistency":"89.08","Distribution":"4.86","Open":"45.54","Plausibility":"84.27","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":26,"model":"45","metrics":{"Accuracy":"60.93","Binary":"77.83","Consistency":"90.3","Distribution":"5.74","Open":"46.01","Plausibility":"84.69","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":27,"model":"IQA (single)","metrics":{"Accuracy":"60.89","Binary":"78.07","Consistency":"93.02","Distribution":"5.31","Open":"45.73","Plausibility":"84.05","Validity":"96.0"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":28,"model":"Ensemble10","metrics":{"Accuracy":"60.87","Binary":"79.12","Consistency":"92.61","Distribution":"8.56","Open":"44.76","Plausibility":"85.63","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":29,"model":"Meta Module, Single","metrics":{"Accuracy":"60.83","Binary":"78.9","Consistency":"92.49","Distribution":"5.54","Open":"44.89","Plausibility":"84.55","Validity":"96.19"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":30,"model":"xpj","metrics":{"Accuracy":"60.7","Binary":"77.41","Consistency":"89.65","Distribution":"6.09","Open":"45.96","Plausibility":"84.55","Validity":"96.37"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":31,"model":"fbe20v3.json","metrics":{"Accuracy":"60.67","Binary":"78.02","Consistency":"89.81","Distribution":"6.41","Open":"45.36","Plausibility":"84.84","Validity":"96.31"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":32,"model":"LININ","metrics":{"Accuracy":"60.59","Binary":"78.44","Consistency":"92.66","Distribution":"7.28","Open":"44.83","Plausibility":"85.38","Validity":"96.57"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":33,"model":"prompt IMT-16","metrics":{"Accuracy":"60.51","Binary":"76.87","Consistency":"88.2","Distribution":"8.48","Open":"46.06","Plausibility":"85.19","Validity":"96.15"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"vv69","metrics":{"Accuracy":"60.42","Binary":"77.12","Consistency":"89.69","Distribution":"6.03","Open":"45.68","Plausibility":"84.56","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":35,"model":"bert_v1","metrics":{"Accuracy":"60.37","Binary":"77.09","Consistency":"89.77","Distribution":"6.43","Open":"45.61","Plausibility":"84.56","Validity":"96.22"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":36,"model":"LXR955, Single Model","metrics":{"Accuracy":"60.33","Binary":"77.16","Consistency":"89.59","Distribution":"5.69","Open":"45.47","Plausibility":"84.53","Validity":"96.35"},"uses_additional_data":false,"paper_date":"2019-08-20","paper":"/paper/lxmert-learning-cross-modality-encoder","paper_url":"https://arxiv.org/abs/1908.07490v3","paper_title":"LXMERT: Learning Cross-Modality Encoder Representations from Transformers","code":"https://github.com/huggingface/transformers","n_code_links":9,"syntology":{"n_ran":4,"n_unverified":11,"n_samples":15,"n_pointer_only_licence":3}},{"rank_in_archive_order":37,"model":"IIE_Morningstar","metrics":{"Accuracy":"60.28","Binary":"77.13","Consistency":"89.47","Distribution":"5.38","Open":"45.41","Plausibility":"84.45","Validity":"96.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":38,"model":"full_nsp_ft_results_submit_predict.json","metrics":{"Accuracy":"60.27","Binary":"76.99","Consistency":"90.16","Distribution":"5.39","Open":"45.51","Plausibility":"84.49","Validity":"96.31"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":39,"model":"TESTOVQA007","metrics":{"Accuracy":"60.18","Binary":"76.97","Consistency":"89.65","Distribution":"5.29","Open":"45.36","Plausibility":"84.47","Validity":"96.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":40,"model":"test gqa","metrics":{"Accuracy":"60.18","Binary":"76.84","Consistency":"89.77","Distribution":"5.65","Open":"45.48","Plausibility":"84.6","Validity":"96.37"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":41,"model":"Future_Test_team","metrics":{"Accuracy":"60.17","Binary":"77.19","Consistency":"89.61","Distribution":"5.83","Open":"45.14","Plausibility":"84.46","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":42,"model":"tmp","metrics":{"Accuracy":"60.14","Binary":"77.15","Consistency":"89.58","Distribution":"5.81","Open":"45.12","Plausibility":"84.47","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":43,"model":"Inspur","metrics":{"Accuracy":"60.07","Binary":"76.84","Consistency":"89.32","Distribution":"6.21","Open":"45.27","Plausibility":"84.55","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":44,"model":"full_nsp_mlm_ft_joint_results_submit_predict.json","metrics":{"Accuracy":"60.02","Binary":"76.37","Consistency":"90.05","Distribution":"5.63","Open":"45.59","Plausibility":"84.34","Validity":"96.29"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":45,"model":"SSRP","metrics":{"Accuracy":"60.01","Binary":"76.77","Consistency":"89.17","Distribution":"6.28","Open":"45.21","Plausibility":"84.46","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":46,"model":"Musan","metrics":{"Accuracy":"59.93","Binary":"79.09","Consistency":"93.72","Distribution":"10.1","Open":"43.02","Plausibility":"85.92","Validity":"96.41"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":47,"model":"gaochongyang9","metrics":{"Accuracy":"59.84","Binary":"76.79","Consistency":"89.52","Distribution":"6.06","Open":"44.89","Plausibility":"84.72","Validity":"96.2"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":48,"model":"PVR","metrics":{"Accuracy":"59.81","Binary":"78.02","Consistency":"91.43","Distribution":"6.0","Open":"43.75","Plausibility":"84.77","Validity":"96.5"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":49,"model":"BgTest","metrics":{"Accuracy":"59.8","Binary":"76.74","Consistency":"89.14","Distribution":"5.11","Open":"44.85","Plausibility":"84.2","Validity":"96.23"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":50,"model":"DAM","metrics":{"Accuracy":"59.72","Binary":"77.97","Consistency":"89.43","Distribution":"6.25","Open":"43.61","Plausibility":"84.89","Validity":"96.55"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":51,"model":"DL16","metrics":{"Accuracy":"59.54","Binary":"77.98","Consistency":"89.21","Distribution":"6.01","Open":"43.26","Plausibility":"84.94","Validity":"96.24"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":52,"model":"mcmi","metrics":{"Accuracy":"59.43","Binary":"77.11","Consistency":"89.05","Distribution":"6.39","Open":"43.82","Plausibility":"84.94","Validity":"96.56"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":53,"model":"rishabh_test","metrics":{"Accuracy":"59.37","Binary":"77.53","Consistency":"88.63","Distribution":"6.06","Open":"43.35","Plausibility":"84.71","Validity":"96.18"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":54,"model":"UNITER + MAC + Graph Networks","metrics":{"Accuracy":"59.29","Binary":"77.31","Consistency":"88.94","Distribution":"5.8","Open":"43.38","Plausibility":"84.43","Validity":"96.3"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":55,"model":"LXMERT-S","metrics":{"Accuracy":"59.12","Binary":"76.69","Consistency":"88.9","Distribution":"5.6","Open":"43.6","Plausibility":"84.78","Validity":"96.43"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":56,"model":"QGCRGN","metrics":{"Accuracy":"59.06","Binary":"76.07","Consistency":"89.81","Distribution":"6.14","Open":"44.04","Plausibility":"82.76","Validity":"93.82"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":57,"model":"gbert1","metrics":{"Accuracy":"58.91","Binary":"76.08","Consistency":"89.52","Distribution":"6.93","Open":"43.75","Plausibility":"84.52","Validity":"96.18"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":58,"model":"glimple_all","metrics":{"Accuracy":"58.88","Binary":"75.07","Consistency":"84.64","Distribution":"5.54","Open":"44.58","Plausibility":"84.86","Validity":"96.23"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":59,"model":"ours-4-gqa_el_tag_v4__pretrain_rel_tag_dist_tc_v7_checkpoint-47-157510-best-4.json","metrics":{"Accuracy":"58.72","Binary":"76.4","Consistency":"89.58","Distribution":"6.58","Open":"43.11","Plausibility":"84.68","Validity":"96.21"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":60,"model":"Partial-MSP","metrics":{"Accuracy":"58.42","Binary":"77.39","Consistency":"90.29","Distribution":"7.86","Open":"41.67","Plausibility":"84.53","Validity":"95.57"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":61,"model":"UCAS-SARI","metrics":{"Accuracy":"58.2","Binary":"75.91","Consistency":"88.25","Distribution":"5.81","Open":"42.57","Plausibility":"84.72","Validity":"96.08"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":62,"model":"stu09e","metrics":{"Accuracy":"58.12","Binary":"76.39","Consistency":"88.01","Distribution":"5.65","Open":"42.0","Plausibility":"84.8","Validity":"96.06"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":63,"model":"happyTeam","metrics":{"Accuracy":"58.06","Binary":"76.6","Consistency":"90.96","Distribution":"7.6","Open":"41.7","Plausibility":"85.27","Validity":"96.31"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":64,"model":"graphRepresentation， Single","metrics":{"Accuracy":"57.89","Binary":"74.54","Consistency":"85.45","Distribution":"5.73","Open":"43.19","Plausibility":"84.99","Validity":"96.4"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":65,"model":"VqaStar-UCAS-SARI","metrics":{"Accuracy":"57.79","Binary":"75.37","Consistency":"88.3","Distribution":"5.65","Open":"42.26","Plausibility":"84.85","Validity":"96.11"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":66,"model":"REX","metrics":{"Accuracy":"57.77","Binary":"75.78","Consistency":"86.85","Distribution":"5.36","Open":"41.86","Plausibility":"84.97","Validity":"96.44"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":67,"model":"MLVQA (single)","metrics":{"Accuracy":"57.65","Binary":"75.22","Consistency":"87.35","Distribution":"5.48","Open":"42.14","Plausibility":"84.73","Validity":"96.18"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":68,"model":"rsa-14word","metrics":{"Accuracy":"57.35","Binary":"75.07","Consistency":"87.61","Distribution":"5.94","Open":"41.71","Plausibility":"84.5","Validity":"95.86"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":69,"model":"result_run_2647872_epoch11","metrics":{"Accuracy":"57.21","Binary":"74.46","Consistency":"87.6","Distribution":"5.6","Open":"41.99","Plausibility":"84.87","Validity":"96.2"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":70,"model":"DeeTee","metrics":{"Accuracy":"57.14","Binary":"75.07","Consistency":"87.36","Distribution":"5.29","Open":"41.31","Plausibility":"84.49","Validity":"95.87"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":71,"model":"BAN","metrics":{"Accuracy":"57.1","Binary":"76.0","Consistency":"91.7","Distribution":"10.52","Open":"40.41","Plausibility":"85.58","Validity":"96.16"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":72,"model":"LCGN","metrics":{"Accuracy":"57.07","Binary":"73.77","Consistency":"84.68","Distribution":"4.7","Open":"42.33","Plausibility":"84.81","Validity":"96.48"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":73,"model":"RSN (Single Model)","metrics":{"Accuracy":"57.01","Binary":"74.78","Consistency":"87.74","Distribution":"6.06","Open":"41.32","Plausibility":"84.25","Validity":"96.03"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":74,"model":"GM6_9_2_train","metrics":{"Accuracy":"56.96","Binary":"74.97","Consistency":"85.12","Distribution":"7.13","Open":"41.06","Plausibility":"84.85","Validity":"96.38"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":75,"model":"wcf-fight","metrics":{"Accuracy":"56.95","Binary":"75.01","Consistency":"90.49","Distribution":"9.5","Open":"41.02","Plausibility":"85.46","Validity":"96.37"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":76,"model":"total14","metrics":{"Accuracy":"56.95","Binary":"74.62","Consistency":"87.71","Distribution":"5.81","Open":"41.36","Plausibility":"84.57","Validity":"95.98"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":77,"model":"Testify","metrics":{"Accuracy":"56.65","Binary":"73.65","Consistency":"84.35","Distribution":"6.07","Open":"41.64","Plausibility":"84.37","Validity":"95.94"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":78,"model":"F205","metrics":{"Accuracy":"56.59","Binary":"73.0","Consistency":"84.7","Distribution":"4.68","Open":"42.11","Plausibility":"84.86","Validity":"96.4"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":79,"model":"Feb_ft2_mergeadd_weightalllstm_picklocw_box5_prep","metrics":{"Accuracy":"56.38","Binary":"74.84","Consistency":"91.71","Distribution":"6.32","Open":"40.09","Plausibility":"83.76","Validity":"95.43"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":80,"model":"MMT-VQA","metrics":{"Accuracy":"56.28","Binary":"73.73","Consistency":"86.86","Distribution":"5.78","Open":"40.87","Plausibility":"84.2","Validity":"96.01"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":81,"model":"IWantADonut","metrics":{"Accuracy":"56.18","Binary":"72.84","Consistency":"85.46","Distribution":"5.42","Open":"41.47","Plausibility":"84.04","Validity":"96.18"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":82,"model":"GIN","metrics":{"Accuracy":"56.16","Binary":"73.56","Consistency":"84.99","Distribution":"5.87","Open":"40.8","Plausibility":"84.83","Validity":"96.4"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":83,"model":"LOGNet+VLR","metrics":{"Accuracy":"56.11","Binary":"72.65","Consistency":"85.51","Distribution":"5.42","Open":"41.52","Plausibility":"84.36","Validity":"96.25"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":84,"model":"Improved SNMN","metrics":{"Accuracy":"56.09","Binary":"73.4","Consistency":"85.11","Distribution":"5.14","Open":"40.82","Plausibility":"84.79","Validity":"96.37"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":85,"model":"ST_VQA","metrics":{"Accuracy":"56.0","Binary":"73.9","Consistency":"87.16","Distribution":"6.02","Open":"40.2","Plausibility":"84.45","Validity":"96.01"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":86,"model":"RD","metrics":{"Accuracy":"55.93","Binary":"71.81","Consistency":"83.2","Distribution":"6.05","Open":"41.93","Plausibility":"85.09","Validity":"96.01"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":87,"model":"Deepblue_Semantics","metrics":{"Accuracy":"55.7","Binary":"72.88","Consistency":"83.52","Distribution":"5.32","Open":"40.53","Plausibility":"84.81","Validity":"96.39"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":88,"model":"LW","metrics":{"Accuracy":"55.65","Binary":"72.86","Consistency":"89.18","Distribution":"9.69","Open":"40.46","Plausibility":"85.27","Validity":"96.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":89,"model":"RSN (Single Model)_v6","metrics":{"Accuracy":"55.57","Binary":"72.39","Consistency":"83.32","Distribution":"10.18","Open":"40.74","Plausibility":"84.24","Validity":"96.15"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":90,"model":"nogg","metrics":{"Accuracy":"55.41","Binary":"72.87","Consistency":"83.06","Distribution":"5.48","Open":"39.99","Plausibility":"84.74","Validity":"96.35"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":91,"model":"abc_test","metrics":{"Accuracy":"55.35","Binary":"72.65","Consistency":"84.17","Distribution":"5.22","Open":"40.08","Plausibility":"84.56","Validity":"96.32"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":92,"model":"KU","metrics":{"Accuracy":"55.0","Binary":"72.09","Consistency":"83.47","Distribution":"5.29","Open":"39.92","Plausibility":"84.66","Validity":"96.34"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":93,"model":"Eden_test","metrics":{"Accuracy":"54.94","Binary":"71.7","Consistency":"82.71","Distribution":"5.1","Open":"40.14","Plausibility":"84.78","Validity":"96.4"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":94,"model":"HDU_ZWF","metrics":{"Accuracy":"54.79","Binary":"72.42","Consistency":"86.1","Distribution":"6.01","Open":"39.23","Plausibility":"84.55","Validity":"95.92"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":95,"model":"vips","metrics":{"Accuracy":"54.15","Binary":"69.3","Consistency":"82.36","Distribution":"5.41","Open":"40.79","Plausibility":"85.15","Validity":"95.99"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":96,"model":"MAC","metrics":{"Accuracy":"54.06","Binary":"71.23","Consistency":"81.59","Distribution":"5.34","Open":"38.91","Plausibility":"84.48","Validity":"96.16"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":97,"model":"5TMT-qe+o","metrics":{"Accuracy":"53.89","Binary":"72.52","Consistency":"87.47","Distribution":"8.66","Open":"37.44","Plausibility":"85.05","Validity":"96.39"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":98,"model":"ZhaoLab","metrics":{"Accuracy":"53.85","Binary":"68.44","Consistency":"80.2","Distribution":"5.84","Open":"40.97","Plausibility":"85.19","Validity":"96.28"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":99,"model":"test","metrics":{"Accuracy":"53.57","Binary":"70.15","Consistency":"81.14","Distribution":"5.32","Open":"38.94","Plausibility":"84.67","Validity":"96.36"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":100,"model":"Sorbonne","metrics":{"Accuracy":"53.31","Binary":"70.41","Consistency":"80.33","Distribution":"6.4","Open":"38.23","Plausibility":"84.32","Validity":"95.99"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":101,"model":"UJCNN","metrics":{"Accuracy":"52.3","Binary":"68.46","Consistency":"84.36","Distribution":"12.54","Open":"38.04","Plausibility":"85.2","Validity":"96.2"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":102,"model":"MJ","metrics":{"Accuracy":"52.19","Binary":"69.15","Consistency":"78.34","Distribution":"5.69","Open":"37.22","Plausibility":"83.44","Validity":"95.45"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":103,"model":"mac_qin","metrics":{"Accuracy":"52.02","Binary":"67.35","Consistency":"80.44","Distribution":"5.64","Open":"38.5","Plausibility":"83.94","Validity":"95.75"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":104,"model":"Mithrandir","metrics":{"Accuracy":"51.87","Binary":"67.99","Consistency":"80.2","Distribution":"6.77","Open":"37.64","Plausibility":"84.35","Validity":"96.25"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":105,"model":"happy","metrics":{"Accuracy":"51.51","Binary":"67.82","Consistency":"79.7","Distribution":"6.16","Open":"37.11","Plausibility":"83.69","Validity":"95.82"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":106,"model":"Space Cat","metrics":{"Accuracy":"51.22","Binary":"69.36","Consistency":"82.44","Distribution":"6.45","Open":"35.2","Plausibility":"83.82","Validity":"96.12"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":107,"model":"BottomUp","metrics":{"Accuracy":"49.74","Binary":"66.64","Consistency":"78.71","Distribution":"5.98","Open":"34.83","Plausibility":"84.57","Validity":"96.18"},"uses_additional_data":false,"paper_date":"2017-07-25","paper":"/paper/bottom-up-and-top-down-attention-for-image","paper_url":"http://arxiv.org/abs/1707.07998v3","paper_title":"Bottom-Up and Top-Down Attention for Image Captioning and Visual Question Answering","code":"https://github.com/facebookresearch/mmf","n_code_links":65,"syntology":{"n_ran":9,"n_unverified":0,"n_samples":9,"n_pointer_only_licence":6}},{"rank_in_archive_order":108,"model":"RAM_BUGGY","metrics":{"Accuracy":"49.28","Binary":"67.59","Consistency":"83.68","Distribution":"14.28","Open":"33.12","Plausibility":"83.41","Validity":"94.95"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":109,"model":"sparsemax15","metrics":{"Accuracy":"49.27","Binary":"66.57","Consistency":"78.51","Distribution":"6.91","Open":"34.0","Plausibility":"84.58","Validity":"95.78"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":110,"model":"mfb+bert","metrics":{"Accuracy":"48.97","Binary":"63.85","Consistency":"83.85","Distribution":"13.72","Open":"35.83","Plausibility":"83.93","Validity":"95.62"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":111,"model":"RES","metrics":{"Accuracy":"48.44","Binary":"65.02","Consistency":"81.19","Distribution":"17.79","Open":"33.81","Plausibility":"85.29","Validity":"96.15"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":112,"model":"LAS","metrics":{"Accuracy":"47.72","Binary":"66.28","Consistency":"84.16","Distribution":"19.05","Open":"31.34","Plausibility":"84.52","Validity":"95.45"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":113,"model":"test","metrics":{"Accuracy":"47.38","Binary":"58.76","Consistency":"73.71","Distribution":"6.29","Open":"37.34","Plausibility":"81.75","Validity":"94.55"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":114,"model":"LSTM+CNN","metrics":{"Accuracy":"46.55","Binary":"63.26","Consistency":"74.57","Distribution":"7.46","Open":"31.8","Plausibility":"84.25","Validity":"96.02"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":115,"model":"113","metrics":{"Accuracy":"45.86","Binary":"64.74","Consistency":"70.57","Distribution":"8.38","Open":"29.2","Plausibility":"86.13","Validity":"96.61"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":116,"model":"Ediburgh-Mila-UCLA","metrics":{"Accuracy":"44.06","Binary":"57.57","Consistency":"38.18","Distribution":"8.35","Open":"32.13","Plausibility":"75.19","Validity":"85.94"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":117,"model":"bear","metrics":{"Accuracy":"43.84","Binary":"59.24","Consistency":"67.71","Distribution":"10.99","Open":"30.24","Plausibility":"84.01","Validity":"95.32"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":118,"model":"CHAIR","metrics":{"Accuracy":"42.75","Binary":"61.21","Consistency":"63.51","Distribution":"7.63","Open":"26.45","Plausibility":"84.2","Validity":"95.99"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":119,"model":"MReaL","metrics":{"Accuracy":"41.63","Binary":"55.12","Consistency":"82.21","Distribution":"13.01","Open":"29.73","Plausibility":"77.4","Validity":"92.27"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":120,"model":"LSTM","metrics":{"Accuracy":"41.07","Binary":"61.9","Consistency":"68.68","Distribution":"17.93","Open":"22.69","Plausibility":"87.3","Validity":"96.39"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":121,"model":"Academia Sinica","metrics":{"Accuracy":"40.3","Binary":"61.18","Consistency":"74.11","Distribution":"40.44","Open":"21.88","Plausibility":"86.13","Validity":"96.14"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":122,"model":"Fj","metrics":{"Accuracy":"37.03","Binary":"56.61","Consistency":"63.96","Distribution":"28.4","Open":"19.74","Plausibility":"85.12","Validity":"95.76"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":123,"model":"Mycsulb","metrics":{"Accuracy":"36.75","Binary":"55.24","Consistency":"69.93","Distribution":"40.84","Open":"20.44","Plausibility":"84.13","Validity":"95.1"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":124,"model":"LocalPrior","metrics":{"Accuracy":"31.24","Binary":"47.9","Consistency":"54.04","Distribution":"13.98","Open":"16.66","Plausibility":"84.31","Validity":"84.33"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":125,"model":"GlobalPrior","metrics":{"Accuracy":"28.9","Binary":"42.94","Consistency":"51.69","Distribution":"93.08","Open":"16.62","Plausibility":"74.81","Validity":"88.86"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":126,"model":"muc_ai","metrics":{"Accuracy":"26.45","Binary":"45.69","Consistency":"55.23","Distribution":"11.49","Open":"9.47","Plausibility":"50.93","Validity":"60.81"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":127,"model":"CNN","metrics":{"Accuracy":"17.82","Binary":"36.05","Consistency":"62.4","Distribution":"19.99","Open":"1.74","Plausibility":"34.84","Validity":"35.78"},"uses_additional_data":false,"paper_date":null,"paper":null,"paper_url":null,"paper_title":"","code":null,"n_code_links":0,"syntology":null}],"since_archive":{"claim":"Results that newer papers report for their own method, placed here by Syntology. A model pointed at the cell in the paper's own table; the number was read from that cell and checked against this leaderboard's metric, dataset, split and scale; an independent check that saw this leaderboard's other rows and every other leaderboard on the same dataset accepted it. Not reviewed by the paper's authors or by the archive's editors, and not ranked against the archive rows.","extraction_file_present":true,"measurement":{"test_papers":883,"papers_with_output":881,"judged_true":108,"judged":110,"wilson95_lower":0.9361,"measured_on":"2026-09-24","frozen_commit":"0e3de0df94"},"measurement_note":"blind adjudication of accepted entries on a held-out split of archive papers, rules frozen before the test","coverage":{"sentence":"Syntology has checked 6,885 of the 9,623 papers on this site that are newer than the archive; results from the others appear after they are checked.","complete":false,"papers_newer_than_archive":9623,"papers_checked":6885,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2737},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-25T09:33:49+00:00","claim":"Per row: N of M harvested code samples from that row's paper executed on a synthesized fixture; the other M-N are unverified. Not a reproduction of the row's number; not a correctness claim. n_pointer_only_licence counts samples the site points at rather than redistributes (a licence axis, independent of ran/unverified).","rows_with_graph_line":4,"rows_with_any_sample_ran":4,"distinct_papers_with_graph_line":3,"distinct_papers_with_any_sample_ran":3,"samples_over_distinct_papers":{"n_ran":15,"n_unverified":11,"n_samples":26,"n_pointer_only_licence":10,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":19,"n_unverified":22,"n_samples":41,"n_pointer_only_licence":13,"note":"row-weighted: a paper behind several rows is counted once per row; inflated relative to samples_over_distinct_papers by design, kept for readers summing the per-row syntology blocks"}}}