{"url":"/sota/question-answering-on-hotpotqa","task":{"name":"Question Answering","url":"/task/question-answering","note":null},"dataset":{"name":"HotpotQA","url":"/dataset/hotpotqa"},"category":"Natural Language Processing","categories":["Miscellaneous","Natural Language Processing","Reasoning"],"category_note":null,"description":"Question answering can be segmented into domain-specific tasks like community question answering and knowledge-base question answering. Popular benchmark datasets for evaluation question answering systems include [SQuAD](/dataset/squad), [HotPotQA](/dataset/hotpotqa), [bAbI](/dataset/babi-1), [TriviaQA](/dataset/triviaqa), [WikiQA](/dataset/wikiqa), and many others. Models for question answering are typically evaluated on metrics like EM and F1. Some recent top performing models are T5 and XLNet.\r\n\r\n( Image credit: [SQuAD](https://rajpurkar.github.io/mlx/qa-and-squad/) )","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":["JOINT-F1","ANS-EM","ANS-F1","SUP-EM","SUP-F1","JOINT-EM"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"JOINT-F1":"higher","ANS-EM":null,"ANS-F1":"higher","SUP-EM":null,"SUP-F1":"higher","JOINT-EM":null}},"counts":{"rows":72,"rows_with_code":18,"rows_with_paper_page":23,"rows_dated":23,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Beam Retrieval","metrics":{"ANS-EM":"0.727","ANS-F1":"0.850","JOINT-EM":"0.505","JOINT-F1":"0.775","SUP-EM":"0.663","SUP-F1":"0.901"},"uses_additional_data":false,"paper_date":"2023-08-17","paper":"/paper/beam-retrieval-general-end-to-end-retrieval","paper_url":"https://arxiv.org/abs/2308.08973v2","paper_title":"End-to-End Beam Retrieval for Multi-Hop Question Answering","code":"https://github.com/ShayekhBinIslam/openrag","n_code_links":3,"syntology":{"n_ran":10,"n_unverified":3,"n_samples":13,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"BigBird-etc","metrics":{"ANS-F1":"0.755","JOINT-F1":"0.736","SUP-F1":"0.891"},"uses_additional_data":false,"paper_date":"2020-07-28","paper":"/paper/big-bird-transformers-for-longer-sequences","paper_url":"https://arxiv.org/abs/2007.14062v2","paper_title":"Big Bird: Transformers for Longer Sequences","code":"https://github.com/huggingface/transformers","n_code_links":14,"syntology":{"n_ran":10,"n_unverified":5,"n_samples":15,"n_pointer_only_licence":11}},{"rank_in_archive_order":3,"model":"AISO","metrics":{"ANS-EM":"0.675","ANS-F1":"0.805","JOINT-EM":"0.449","JOINT-F1":"0.720","SUP-EM":"0.612","SUP-F1":"0.860"},"uses_additional_data":false,"paper_date":"2021-09-14","paper":"/paper/adaptive-information-seeking-for-open-domain","paper_url":"https://arxiv.org/abs/2109.06747v1","paper_title":"Adaptive Information Seeking for Open-Domain Question Answering","code":"https://github.com/zycdev/aiso","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"Chain-of-Skills","metrics":{"ANS-EM":"0.674","ANS-F1":"0.801","JOINT-EM":"0.457","JOINT-F1":"0.717","SUP-EM":"0.613","SUP-F1":"0.853"},"uses_additional_data":false,"paper_date":"2023-05-04","paper":"/paper/chain-of-skills-a-configurable-model-for-open","paper_url":"https://arxiv.org/abs/2305.03130v2","paper_title":"Chain-of-Skills: A Configurable Model for Open-domain Question Answering","code":"https://github.com/mayer123/udt-qa","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"TPRR","metrics":{"ANS-EM":"0.670","ANS-F1":"0.795","JOINT-EM":"0.444","JOINT-F1":"0.708","SUP-EM":"0.594","SUP-F1":"0.843"},"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":"HopRetriever + Sp-search","metrics":{"ANS-EM":"0.671","ANS-F1":"0.799","JOINT-EM":"0.432","JOINT-F1":"0.706","SUP-EM":"0.574","SUP-F1":"0.835"},"uses_additional_data":false,"paper_date":"2020-12-31","paper":"/paper/hopretriever-retrieve-hops-over-wikipedia-to","paper_url":"https://arxiv.org/abs/2012.15534v1","paper_title":"HopRetriever: Retrieve Hops over Wikipedia to Answer Complex Questions","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"EBS-Large","metrics":{"ANS-EM":"0.662","ANS-F1":"0.793","JOINT-EM":"0.420","JOINT-F1":"0.700","SUP-EM":"0.573","SUP-F1":"0.840"},"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":"HopRetriever","metrics":{"ANS-EM":"0.671","ANS-F1":"0.799","JOINT-EM":"0.431","JOINT-F1":"0.698","SUP-EM":"0.572","SUP-F1":"0.826"},"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":"IRRR+","metrics":{"ANS-EM":"0.663","ANS-F1":"0.791","JOINT-EM":"0.428","JOINT-F1":"0.696","SUP-EM":"0.569","SUP-F1":"0.832"},"uses_additional_data":false,"paper_date":"2020-10-23","paper":"/paper/retrieve-rerank-read-then-iterate-answering","paper_url":"https://arxiv.org/abs/2010.12527v4","paper_title":"Answering Open-Domain Questions of Varying Reasoning Steps from Text","code":"https://github.com/beerqa/irrr","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":10,"model":"EBS-SH","metrics":{"ANS-EM":"0.655","ANS-F1":"0.786","JOINT-EM":"0.409","JOINT-F1":"0.689","SUP-EM":"0.559","SUP-F1":"0.831"},"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":11,"model":"IRRR","metrics":{"ANS-EM":"0.657","ANS-F1":"0.782","JOINT-EM":"0.421","JOINT-F1":"0.686","SUP-EM":"0.559","SUP-F1":"0.821"},"uses_additional_data":false,"paper_date":"2020-10-23","paper":"/paper/retrieve-rerank-read-then-iterate-answering","paper_url":"https://arxiv.org/abs/2010.12527v4","paper_title":"Answering Open-Domain Questions of Varying Reasoning Steps from Text","code":"https://github.com/beerqa/irrr","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":0,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":12,"model":"HopRetriever-V2","metrics":{"ANS-EM":"0.648","ANS-F1":"0.778","JOINT-EM":"0.410","JOINT-F1":"0.678","SUP-EM":"0.561","SUP-F1":"0.818"},"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":"AFSGraph-retriever","metrics":{"ANS-EM":"0.646","ANS-F1":"0.778","JOINT-EM":"0.411","JOINT-F1":"0.670","SUP-EM":"0.557","SUP-F1":"0.812"},"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":"Recursive Dense Retriever","metrics":{"ANS-EM":"0.623","ANS-F1":"0.753","JOINT-EM":"0.418","JOINT-F1":"0.666","SUP-EM":"0.575","SUP-F1":"0.809"},"uses_additional_data":false,"paper_date":"2020-09-27","paper":"/paper/answering-complex-open-domain-questions-with","paper_url":"https://arxiv.org/abs/2009.12756v2","paper_title":"Answering Complex Open-Domain Questions with Multi-Hop Dense Retrieval","code":"https://github.com/facebookresearch/multihop_dense_retrieval","n_code_links":1,"syntology":null},{"rank_in_archive_order":15,"model":"Step-by-Step Retriever","metrics":{"ANS-EM":"0.630","ANS-F1":"0.754","JOINT-EM":"0.404","JOINT-F1":"0.662","SUP-EM":"0.546","SUP-F1":"0.800"},"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":16,"model":"DDRQA","metrics":{"ANS-EM":"0.625","ANS-F1":"0.759","JOINT-EM":"0.360","JOINT-F1":"0.639","SUP-EM":"0.510","SUP-F1":"0.789"},"uses_additional_data":false,"paper_date":"2020-09-16","paper":"/paper/ddrqa-dynamic-document-reranking-for-open","paper_url":"https://arxiv.org/abs/2009.07465v5","paper_title":"Answering Any-hop Open-domain Questions with Iterative Document Reranking","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":17,"model":"HopRetriever-V1","metrics":{"ANS-EM":"0.608","ANS-F1":"0.739","JOINT-EM":"0.380","JOINT-F1":"0.639","SUP-EM":"0.531","SUP-F1":"0.793"},"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":"DR model large","metrics":{"ANS-EM":"0.620","ANS-F1":"0.753","JOINT-EM":"0.354","JOINT-F1":"0.630","SUP-EM":"0.499","SUP-F1":"0.778"},"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":"Model name","metrics":{"ANS-EM":"0.617","ANS-F1":"0.746","JOINT-EM":"0.368","JOINT-F1":"0.629","SUP-EM":"0.500","SUP-F1":"0.772"},"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":20,"model":"HopAns","metrics":{"ANS-EM":"0.617","ANS-F1":"0.746","JOINT-EM":"0.368","JOINT-F1":"0.629","SUP-EM":"0.500","SUP-F1":"0.772"},"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":"Anonymous","metrics":{"ANS-EM":"0.604","ANS-F1":"0.732","JOINT-EM":"0.380","JOINT-F1":"0.629","SUP-EM":"0.520","SUP-F1":"0.771"},"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":"Multi-dimensional-AFSGraph","metrics":{"ANS-EM":"0.615","ANS-F1":"0.746","JOINT-EM":"0.362","JOINT-F1":"0.624","SUP-EM":"0.503","SUP-F1":"0.772"},"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":"HGN-albert + SemanticRetrievalMRS IR","metrics":{"ANS-EM":"0.597","ANS-F1":"0.714","JOINT-EM":"0.379","JOINT-F1":"0.623","SUP-EM":"0.510","SUP-F1":"0.774"},"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":"Tree-shaped-cluster","metrics":{"ANS-EM":"0.603","ANS-F1":"0.731","JOINT-EM":"0.359","JOINT-F1":"0.617","SUP-EM":"0.499","SUP-F1":"0.768"},"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":"AFSgraph","metrics":{"ANS-EM":"0.601","ANS-F1":"0.730","JOINT-EM":"0.359","JOINT-F1":"0.617","SUP-EM":"0.500","SUP-F1":"0.769"},"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":"Robustly Fine-tuned Graph-based Recurrent Retriever","metrics":{"ANS-EM":"0.600","ANS-F1":"0.730","JOINT-EM":"0.354","JOINT-F1":"0.612","SUP-EM":"0.491","SUP-F1":"0.764"},"uses_additional_data":false,"paper_date":"2019-11-24","paper":"/paper/learning-to-retrieve-reasoning-paths-over-1","paper_url":"https://arxiv.org/abs/1911.10470v2","paper_title":"Learning to Retrieve Reasoning Paths over Wikipedia Graph for Question Answering","code":"https://github.com/AkariAsai/learning_to_retrieve_reasoning_paths","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":3,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":27,"model":"AFSgraph  model","metrics":{"ANS-EM":"0.601","ANS-F1":"0.730","JOINT-EM":"0.350","JOINT-F1":"0.609","SUP-EM":"0.485","SUP-F1":"0.759"},"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":"HGN-large + SemanticRetrievalMRS IR","metrics":{"ANS-EM":"0.579","ANS-F1":"0.699","JOINT-EM":"0.372","JOINT-F1":"0.607","SUP-EM":"0.510","SUP-F1":"0.768"},"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":"RoBERTa-DenseRetriever-Fast","metrics":{"ANS-EM":"0.598","ANS-F1":"0.727","JOINT-EM":"0.345","JOINT-F1":"0.602","SUP-EM":"0.480","SUP-F1":"0.749"},"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":"DPR-recurrent","metrics":{"ANS-EM":"0.598","ANS-F1":"0.727","JOINT-EM":"0.345","JOINT-F1":"0.602","SUP-EM":"0.480","SUP-F1":"0.749"},"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":"RoBERTa-DenseRetriever","metrics":{"ANS-EM":"0.596","ANS-F1":"0.724","JOINT-EM":"0.345","JOINT-F1":"0.601","SUP-EM":"0.479","SUP-F1":"0.748"},"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":"HGN + SemanticRetrievalMRS IR","metrics":{"ANS-EM":"0.567","ANS-F1":"0.692","JOINT-EM":"0.356","JOINT-F1":"0.599","SUP-EM":"0.500","SUP-F1":"0.764"},"uses_additional_data":false,"paper_date":"2019-11-09","paper":"/paper/hierarchical-graph-network-for-multi-hop","paper_url":"https://arxiv.org/abs/1911.03631v4","paper_title":"Hierarchical Graph Network for Multi-hop Question Answering","code":"https://github.com/yuwfan/HGN","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":3,"n_samples":6,"n_pointer_only_licence":5}},{"rank_in_archive_order":33,"model":"DFGN","metrics":{"JOINT-F1":"0.5982"},"uses_additional_data":false,"paper_date":"2019-05-16","paper":"/paper/dynamically-fused-graph-network-for-multi-hop","paper_url":"https://arxiv.org/abs/1905.06933v3","paper_title":"Dynamically Fused Graph Network for Multi-hop Reasoning","code":"https://github.com/woshiyyya/DFGN-pytorch","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":34,"model":"SAFSR model","metrics":{"ANS-EM":"0.589","ANS-F1":"0.716","JOINT-EM":"0.345","JOINT-F1":"0.598","SUP-EM":"0.480","SUP-F1":"0.757"},"uses_additional_data":false,"paper_date":"2018-09-25","paper":"/paper/hotpotqa-a-dataset-for-diverse-explainable","paper_url":"http://arxiv.org/abs/1809.09600v1","paper_title":"HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering","code":"https://github.com/hotpotqa/hotpot","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":35,"model":"GraphRR-Fast","metrics":{"ANS-EM":"0.582","ANS-F1":"0.709","JOINT-EM":"0.310","JOINT-F1":"0.569","SUP-EM":"0.429","SUP-F1":"0.713"},"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":"DR model","metrics":{"ANS-EM":"0.588","ANS-F1":"0.717","JOINT-EM":"0.293","JOINT-F1":"0.568","SUP-EM":"0.416","SUP-F1":"0.725"},"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":37,"model":"Quark + SemanticRetrievalMRS IR","metrics":{"ANS-EM":"0.555","ANS-F1":"0.675","JOINT-EM":"0.329","JOINT-F1":"0.562","SUP-EM":"0.456","SUP-F1":"0.730"},"uses_additional_data":false,"paper_date":"2020-04-14","paper":"/paper/a-simple-yet-strong-pipeline-for-hotpotqa","paper_url":"https://arxiv.org/abs/2004.06753v1","paper_title":"A Simple Yet Strong Pipeline for HotpotQA","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":38,"model":"GAR-BERT","metrics":{"ANS-EM":"0.523","ANS-F1":"0.648","JOINT-EM":"0.330","JOINT-F1":"0.561","SUP-EM":"0.490","SUP-F1":"0.747"},"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":"Graph-based Recurrent Retriever","metrics":{"ANS-EM":"0.560","ANS-F1":"0.689","JOINT-EM":"0.292","JOINT-F1":"0.553","SUP-EM":"0.441","SUP-F1":"0.730"},"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":"MIR+EPS+BERT","metrics":{"ANS-EM":"0.529","ANS-F1":"0.648","JOINT-EM":"0.312","JOINT-F1":"0.548","SUP-EM":"0.428","SUP-F1":"0.720"},"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":"GAR","metrics":{"ANS-EM":"0.482","ANS-F1":"0.613","JOINT-EM":"0.306","JOINT-F1":"0.530","SUP-EM":"0.483","SUP-F1":"0.739"},"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":"Transformer-XH-final","metrics":{"ANS-EM":"0.516","ANS-F1":"0.641","JOINT-EM":"0.261","JOINT-F1":"0.513","SUP-EM":"0.409","SUP-F1":"0.714"},"uses_additional_data":false,"paper_date":"2020-05-01","paper":"/paper/transformer-xh-multi-evidence-reasoning-with","paper_url":"https://openreview.net/forum?id=r1eIiCNYwS","paper_title":"Transformer-XH: Multi-Evidence Reasoning with eXtra Hop Attention","code":"https://github.com/microsoft/Transformer-XH","n_code_links":1,"syntology":null},{"rank_in_archive_order":43,"model":"Transformer-XH","metrics":{"ANS-EM":"0.490","ANS-F1":"0.608","JOINT-EM":"0.271","JOINT-F1":"0.496","SUP-EM":"0.417","SUP-F1":"0.700"},"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":"SemanticRetrievalMRS","metrics":{"ANS-EM":"0.453","ANS-F1":"0.573","JOINT-EM":"0.251","JOINT-F1":"0.476","SUP-EM":"0.387","SUP-F1":"0.708"},"uses_additional_data":false,"paper_date":"2019-09-17","paper":"/paper/revealing-the-importance-of-semantic","paper_url":"https://arxiv.org/abs/1909.08041v1","paper_title":"Revealing the Importance of Semantic Retrieval for Machine Reading at Scale","code":"https://github.com/easonnie/semanticRetrievalMRS","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":11,"n_samples":17,"n_pointer_only_licence":0}},{"rank_in_archive_order":45,"model":"DrKIT","metrics":{"ANS-EM":"0.421","ANS-F1":"0.517","JOINT-EM":"0.247","JOINT-F1":"0.429","SUP-EM":"0.371","SUP-F1":"0.598"},"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":"Entity-centric BERT Pipeline","metrics":{"ANS-EM":"0.418","ANS-F1":"0.531","JOINT-EM":"0.170","JOINT-F1":"0.392","SUP-EM":"0.263","SUP-F1":"0.573"},"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":"PR-Bert","metrics":{"ANS-EM":"0.433","ANS-F1":"0.538","JOINT-EM":"0.145","JOINT-F1":"0.391","SUP-EM":"0.219","SUP-F1":"0.596"},"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":"GoldEn Retriever","metrics":{"ANS-EM":"0.379","ANS-F1":"0.486","JOINT-EM":"0.180","JOINT-F1":"0.391","SUP-EM":"0.307","SUP-F1":"0.642"},"uses_additional_data":false,"paper_date":"2019-10-15","paper":"/paper/answering-complex-open-domain-questions","paper_url":"https://arxiv.org/abs/1910.07000v1","paper_title":"Answering Complex Open-domain Questions Through Iterative Query Generation","code":"https://github.com/qipeng/golden-retriever","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":49,"model":"SAFSr-Bert","metrics":{"ANS-EM":"0.394","ANS-F1":"0.514","JOINT-EM":"0.133","JOINT-F1":"0.370","SUP-EM":"0.242","SUP-F1":"0.585"},"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":"Cognitive Graph QA","metrics":{"ANS-EM":"0.371","ANS-F1":"0.489","JOINT-EM":"0.124","JOINT-F1":"0.349","SUP-EM":"0.228","SUP-F1":"0.577"},"uses_additional_data":false,"paper_date":"2019-05-14","paper":"/paper/cognitive-graph-for-multi-hop-reading","paper_url":"https://arxiv.org/abs/1905.05460v2","paper_title":"Cognitive Graph for Multi-Hop Reading Comprehension at Scale","code":"https://github.com/THUDM/CogQA","n_code_links":2,"syntology":{"n_ran":5,"n_unverified":4,"n_samples":9,"n_pointer_only_licence":0}},{"rank_in_archive_order":51,"model":"GAR-NOSF","metrics":{"ANS-EM":"0.475","ANS-F1":"0.606","JOINT-EM":"0.049","JOINT-F1":"0.334","SUP-EM":"0.076","SUP-F1":"0.448"},"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":"IKFGraph","metrics":{"ANS-EM":"0.358","ANS-F1":"0.453","JOINT-EM":"0.115","JOINT-F1":"0.304","SUP-EM":"0.160","SUP-F1":"0.512"},"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":"AnonymousQ","metrics":{"ANS-EM":"0.369","ANS-F1":"0.460","JOINT-EM":"0.115","JOINT-F1":"0.291","SUP-EM":"0.153","SUP-F1":"0.468"},"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":"HGN Model-reproduce","metrics":{"ANS-EM":"0.335","ANS-F1":"0.427","JOINT-EM":"0.110","JOINT-F1":"0.284","SUP-EM":"0.156","SUP-F1":"0.493"},"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":"MUPPET","metrics":{"ANS-EM":"0.306","ANS-F1":"0.403","JOINT-EM":"0.109","JOINT-F1":"0.270","SUP-EM":"0.167","SUP-F1":"0.473"},"uses_additional_data":false,"paper_date":"2019-06-15","paper":"/paper/multi-hop-paragraph-retrieval-for-open-domain","paper_url":"https://arxiv.org/abs/1906.06606v1","paper_title":"Multi-Hop Paragraph Retrieval for Open-Domain Question Answering","code":"https://github.com/yairf11/MUPPET","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":5,"n_samples":5,"n_pointer_only_licence":5}},{"rank_in_archive_order":56,"model":"GRN + BERT","metrics":{"ANS-EM":"0.299","ANS-F1":"0.391","JOINT-EM":"0.083","JOINT-F1":"0.258","SUP-EM":"0.132","SUP-F1":"0.497"},"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":"Entity-centric IR","metrics":{"ANS-EM":"0.354","ANS-F1":"0.463","JOINT-EM":"0.000","JOINT-F1":"0.255","SUP-EM":"0.001","SUP-F1":"0.432"},"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":"KGNN","metrics":{"ANS-EM":"0.277","ANS-F1":"0.372","JOINT-EM":"0.070","JOINT-F1":"0.247","SUP-EM":"0.127","SUP-F1":"0.472"},"uses_additional_data":false,"paper_date":"2019-11-06","paper":"/paper/multi-paragraph-reasoning-with-knowledge","paper_url":"https://arxiv.org/abs/1911.02170v1","paper_title":"Multi-Paragraph Reasoning with Knowledge-enhanced Graph Neural Network","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":59,"model":"SAQA","metrics":{"ANS-EM":"0.284","ANS-F1":"0.386","JOINT-EM":"0.086","JOINT-F1":"0.245","SUP-EM":"0.147","SUP-F1":"0.472"},"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":"GRN","metrics":{"ANS-EM":"0.273","ANS-F1":"0.365","JOINT-EM":"0.074","JOINT-F1":"0.236","SUP-EM":"0.122","SUP-F1":"0.488"},"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":"QFE","metrics":{"ANS-EM":"0.287","ANS-F1":"0.381","JOINT-EM":"0.087","JOINT-F1":"0.231","SUP-EM":"0.142","SUP-F1":"0.444"},"uses_additional_data":false,"paper_date":"2019-05-21","paper":"/paper/answering-while-summarizing-multi-task","paper_url":"https://arxiv.org/abs/1905.08511v2","paper_title":"Answering while Summarizing: Multi-task Learning for Multi-hop QA with Evidence Extraction","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":62,"model":"SAFSr_model","metrics":{"ANS-EM":"0.289","ANS-F1":"0.391","JOINT-EM":"0.041","JOINT-F1":"0.209","SUP-EM":"0.080","SUP-F1":"0.406"},"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":"SuppBERT","metrics":{"ANS-EM":"0.236","ANS-F1":"0.320","JOINT-EM":"0.033","JOINT-F1":"0.175","SUP-EM":"0.056","SUP-F1":"0.400"},"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":"Baseline Model","metrics":{"ANS-EM":"0.240","ANS-F1":"0.329","JOINT-EM":"0.019","JOINT-F1":"0.162","SUP-EM":"0.039","SUP-F1":"0.377"},"uses_additional_data":false,"paper_date":"2018-09-25","paper":"/paper/hotpotqa-a-dataset-for-diverse-explainable","paper_url":"http://arxiv.org/abs/1809.09600v1","paper_title":"HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering","code":"https://github.com/hotpotqa/hotpot","n_code_links":2,"syntology":{"n_ran":3,"n_unverified":5,"n_samples":8,"n_pointer_only_licence":0}},{"rank_in_archive_order":65,"model":"tes","metrics":{"ANS-EM":"0.074","ANS-F1":"0.121","JOINT-EM":"0.000","JOINT-F1":"0.011","SUP-EM":"0.000","SUP-F1":"0.078"},"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":"PromptRank-fewshot-2-demo","metrics":{"ANS-EM":"0.581","ANS-F1":"0.711","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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":"graph-recurrent-retriever+roberta-base w. S/R-pretraining","metrics":{"ANS-EM":"0.581","ANS-F1":"0.710","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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":"TPReasoner w/o BERT","metrics":{"ANS-EM":"0.360","ANS-F1":"0.474","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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":"MultiQA","metrics":{"ANS-EM":"0.307","ANS-F1":"0.402","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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":"DecompRC","metrics":{"ANS-EM":"0.300","ANS-F1":"0.407","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"uses_additional_data":false,"paper_date":"2019-06-07","paper":"/paper/multi-hop-reading-comprehension-through","paper_url":"https://arxiv.org/abs/1906.02916v2","paper_title":"Multi-hop Reading Comprehension through Question Decomposition and Rescoring","code":"https://github.com/shmsw25/DecompRC","n_code_links":2,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":71,"model":"","metrics":{"ANS-EM":"0.300","ANS-F1":"0.407","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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":"Mistral multi hop with very large sources","metrics":{"ANS-EM":"0.080","ANS-F1":"0.221","JOINT-EM":"0.000","JOINT-F1":"0.000","SUP-EM":"0.000","SUP-F1":"0.000"},"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 6,795 of the 9,581 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":9581,"papers_checked":6795,"papers_extracted_not_yet_verified":0,"boards_without_verdict":2,"papers_not_yet_extracted":2785},"order":"newest first by month (arXiv date, else the arXiv-id month), then arXiv id descending","columns":[],"entries":[]},"syntology":{"read_at":"2026-09-24T18:15:14+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":14,"rows_with_any_sample_ran":13,"distinct_papers_with_graph_line":12,"distinct_papers_with_any_sample_ran":11,"samples_over_distinct_papers":{"n_ran":46,"n_unverified":41,"n_samples":87,"n_pointer_only_licence":26,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":52,"n_unverified":46,"n_samples":98,"n_pointer_only_licence":29,"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"}}}