{"url":"/sota/3d-dense-captioning-on-scanrefer-dataset","task":{"name":"3D dense captioning","url":"/task/3d-dense-captioning","note":null},"dataset":{"name":"ScanRefer Dataset","url":"/dataset/scanrefer-dataset"},"category":"Computer Vision","categories":["Computer Vision","Natural Language Processing"],"category_note":null,"description":"Dense captioning in 3D point clouds is an emerging vision-and-language task involving object-level 3D scene understanding. Apart from coarse semantic class prediction and bounding box regression as in traditional 3D object detection, 3D dense captioning aims at producing a further and finer instance-level label of natural language description on visual appearance and spatial relations for each scene object of interest.","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":["CIDEr","BLEU-4","METEOR","ROUGE-L"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"CIDEr":null,"BLEU-4":"higher","METEOR":null,"ROUGE-L":"higher"}},"counts":{"rows":12,"rows_with_code":7,"rows_with_paper_page":12,"rows_dated":12,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"3D CoCa","metrics":{"BLEU-4":"45.56","CIDEr":"85.42","METEOR":"30.95","ROUGE-L":"61.98"},"uses_additional_data":false,"paper_date":"2025-04-13","paper":"/paper/3d-coca-contrastive-learners-are-3d","paper_url":"https://arxiv.org/abs/2504.09518v1","paper_title":"3D CoCa: Contrastive Learners are 3D Captioners","code":"https://github.com/AIGeeksGroup/3DCoCa","n_code_links":1,"syntology":null},{"rank_in_archive_order":2,"model":"See It All","metrics":{"BLEU-4":"42.17","CIDEr":"83.14","METEOR":"27.92","ROUGE-L":"59.44"},"uses_additional_data":false,"paper_date":"2024-08-14","paper":"/paper/see-it-all-contextualized-late-aggregation","paper_url":"https://arxiv.org/abs/2408.07648v1","paper_title":"See It All: Contextualized Late Aggregation for 3D Dense Captioning","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":3,"model":"BiCA","metrics":{"BLEU-4":"40.16","CIDEr":"80.14","METEOR":"27.76","ROUGE-L":"56.10"},"uses_additional_data":false,"paper_date":"2024-08-13","paper":"/paper/bi-directional-contextual-attention-for-3d","paper_url":"https://arxiv.org/abs/2408.06662v1","paper_title":"Bi-directional Contextual Attention for 3D Dense Captioning","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"Vote2Cap-DETR++","metrics":{"BLEU-4":"41.37","CIDEr":"76.36","METEOR":"28.70","ROUGE-L":"60.00"},"uses_additional_data":false,"paper_date":"2023-09-06","paper":"/paper/vote2cap-detr-decoupling-localization-and","paper_url":"https://arxiv.org/abs/2309.02999v1","paper_title":"Vote2Cap-DETR++: Decoupling Localization and Describing for End-to-End 3D Dense Captioning","code":"https://github.com/ch3cook-fdu/vote2cap-detr","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"Vote2Cap-DETR","metrics":{"BLEU-4":"39.34","CIDEr":"71.45","METEOR":"28.25","ROUGE-L":"59.33"},"uses_additional_data":false,"paper_date":"2023-01-06","paper":"/paper/end-to-end-3d-dense-captioning-with-vote2cap","paper_url":"https://arxiv.org/abs/2301.02508v1","paper_title":"End-to-End 3D Dense Captioning with Vote2Cap-DETR","code":"https://github.com/ch3cook-fdu/vote2cap-detr","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"3DJCG","metrics":{"BLEU-4":"39.67","CIDEr":"60.86","METEOR":"27.45","ROUGE-L":"59.02"},"uses_additional_data":false,"paper_date":"2022-01-01","paper":"/paper/3djcg-a-unified-framework-for-joint-dense","paper_url":"http://openaccess.thecvf.com//content/CVPR2022/html/Cai_3DJCG_A_Unified_Framework_for_Joint_Dense_Captioning_and_Visual_CVPR_2022_paper.html","paper_title":"3DJCG: A Unified Framework for Joint Dense Captioning and Visual Grounding on 3D Point Clouds","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"MORE","metrics":{"BLEU-4":"35.41","CIDEr":"58.89","METEOR":"26.36","ROUGE-L":"55.41"},"uses_additional_data":false,"paper_date":"2022-03-10","paper":"/paper/more-multi-order-relation-mining-for-dense","paper_url":"https://arxiv.org/abs/2203.05203v2","paper_title":"MORE: Multi-Order RElation Mining for Dense Captioning in 3D Scenes","code":"https://github.com/SxJyJay/MORE","n_code_links":1,"syntology":{"n_ran":6,"n_unverified":3,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"SpaCap3d","metrics":{"BLEU-4":"35.30","CIDEr":"58.06","METEOR":"26.16","ROUGE-L":"55.03"},"uses_additional_data":false,"paper_date":"2022-04-22","paper":"/paper/spatiality-guided-transformer-for-3d-dense","paper_url":"https://arxiv.org/abs/2204.10688v1","paper_title":"Spatiality-guided Transformer for 3D Dense Captioning on Point Clouds","code":"https://github.com/heng-hw/spacap3d","n_code_links":1,"syntology":null},{"rank_in_archive_order":9,"model":"Scan2Cap","metrics":{"BLEU-4":"34.25","CIDEr":"53.73","METEOR":"26.14","ROUGE-L":"54.95"},"uses_additional_data":false,"paper_date":"2020-12-03","paper":"/paper/scan2cap-context-aware-dense-captioning-in","paper_url":"https://arxiv.org/abs/2012.02206v1","paper_title":"Scan2Cap: Context-aware Dense Captioning in RGB-D Scans","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":10,"model":"Contextual","metrics":{"BLEU-4":"26.64","CIDEr":"50.29","METEOR":"22.57","ROUGE-L":"44.71"},"uses_additional_data":false,"paper_date":"2022-10-08","paper":"/paper/contextual-modeling-for-3d-dense-captioning","paper_url":"https://arxiv.org/abs/2210.03925v1","paper_title":"Contextual Modeling for 3D Dense Captioning on Point Clouds","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":11,"model":"3D-VLP","metrics":{"BLEU-4":"31.87","CIDEr":"50.02","METEOR":"24.53","ROUGE-L":"51.17"},"uses_additional_data":false,"paper_date":"2023-01-01","paper":"/paper/context-aware-alignment-and-mutual-masking","paper_url":"http://openaccess.thecvf.com//content/CVPR2023/html/Jin_Context-Aware_Alignment_and_Mutual_Masking_for_3D-Language_Pre-Training_CVPR_2023_paper.html","paper_title":"Context-Aware Alignment and Mutual Masking for 3D-Language Pre-Training","code":"https://github.com/leolyj/3d-vlp","n_code_links":1,"syntology":null},{"rank_in_archive_order":12,"model":"χ-Tran2Cap","metrics":{"BLEU-4":"23.83","CIDEr":"41.52","METEOR":"21.90","ROUGE-L":"44.97"},"uses_additional_data":false,"paper_date":"2022-03-02","paper":"/paper/x-trans2cap-cross-modal-knowledge-transfer","paper_url":"https://arxiv.org/abs/2203.00843v3","paper_title":"X-Trans2Cap: Cross-Modal Knowledge Transfer using Transformer for 3D Dense Captioning","code":"https://github.com/curryyuan/x-trans2cap","n_code_links":1,"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":2,"rows_with_any_sample_ran":2,"distinct_papers_with_graph_line":2,"distinct_papers_with_any_sample_ran":2,"samples_over_distinct_papers":{"n_ran":7,"n_unverified":5,"n_samples":12,"n_pointer_only_licence":0,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":7,"n_unverified":5,"n_samples":12,"n_pointer_only_licence":0,"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"}}}