{"url":"/sota/question-answering-on-triviaqa","task":{"name":"Question Answering","url":"/task/question-answering","note":null},"dataset":{"name":"TriviaQA","url":"/dataset/triviaqa"},"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":["EM","F1"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"EM":null,"F1":"higher"}},"counts":{"rows":56,"rows_with_code":36,"rows_with_paper_page":55,"rows_dated":55,"rows_using_additional_data":13},"rows":[{"rank_in_archive_order":1,"model":"Claude 2 (few-shot, k=5)","metrics":{"EM":"87.5"},"uses_additional_data":false,"paper_date":"2023-07-11","paper":"/paper/model-card-and-evaluations-for-claude-models","paper_url":"https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf","paper_title":"Model Card and Evaluations for Claude Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":2,"model":"GPT-4-0613","metrics":{"EM":"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":3,"model":"Claude 1.3 (few-shot, k=5)","metrics":{"EM":"86.7"},"uses_additional_data":false,"paper_date":"2023-07-11","paper":"/paper/model-card-and-evaluations-for-claude-models","paper_url":"https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf","paper_title":"Model Card and Evaluations for Claude Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":4,"model":"RankRAG-llama3-70b (Zero-Shot, KILT)","metrics":{"EM":"86.5"},"uses_additional_data":true,"paper_date":"2024-07-02","paper":"/paper/rankrag-unifying-context-ranking-with","paper_url":"https://arxiv.org/abs/2407.02485v1","paper_title":"RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":5,"model":"PaLM 2-L (one-shot)","metrics":{"EM":"86.1"},"uses_additional_data":true,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":6,"model":"ChatQA-1.5-llama3-70b (Zero-Shot, KILT)","metrics":{"EM":"85.6"},"uses_additional_data":true,"paper_date":"2024-01-18","paper":"/paper/chatqa-building-gpt-4-level-conversational-qa","paper_url":"https://arxiv.org/abs/2401.10225v5","paper_title":"ChatQA: Surpassing GPT-4 on Conversational QA and RAG","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":7,"model":"LLaMA 2 70B (one-shot)","metrics":{"EM":"85"},"uses_additional_data":false,"paper_date":"2023-07-18","paper":"/paper/llama-2-open-foundation-and-fine-tuned-chat","paper_url":"https://arxiv.org/abs/2307.09288v2","paper_title":"Llama 2: Open Foundation and Fine-Tuned Chat Models","code":"https://github.com/facebookresearch/llama","n_code_links":19,"syntology":{"n_ran":31,"n_unverified":21,"n_samples":52,"n_pointer_only_licence":16}},{"rank_in_archive_order":8,"model":"GPT-4-0613 (Zero-shot)","metrics":{"EM":"84.8"},"uses_additional_data":false,"paper_date":"2023-03-15","paper":"/paper/gpt-4-technical-report-1","paper_url":"https://arxiv.org/abs/2303.08774v5","paper_title":"GPT-4 Technical Report","code":"https://github.com/openai/evals","n_code_links":11,"syntology":{"n_ran":5,"n_unverified":0,"n_samples":5,"n_pointer_only_licence":1}},{"rank_in_archive_order":9,"model":"RankRAG-llama3-8b (Zero-Shot, KILT)","metrics":{"EM":"82.9"},"uses_additional_data":true,"paper_date":"2024-07-02","paper":"/paper/rankrag-unifying-context-ranking-with","paper_url":"https://arxiv.org/abs/2407.02485v1","paper_title":"RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":10,"model":"PaLM 2-M (one-shot)","metrics":{"EM":"81.7"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":11,"model":"PaLM-540B (Few-Shot)","metrics":{"EM":"81.4"},"uses_additional_data":true,"paper_date":"2022-04-05","paper":"/paper/palm-scaling-language-modeling-with-pathways-1","paper_url":"https://arxiv.org/abs/2204.02311v5","paper_title":"PaLM: Scaling Language Modeling with Pathways","code":"https://github.com/lucidrains/CoCa-pytorch","n_code_links":7,"syntology":{"n_ran":32,"n_unverified":5,"n_samples":37,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"PaLM-540B (One-Shot)","metrics":{"EM":"81.4"},"uses_additional_data":false,"paper_date":"2022-04-05","paper":"/paper/palm-scaling-language-modeling-with-pathways-1","paper_url":"https://arxiv.org/abs/2204.02311v5","paper_title":"PaLM: Scaling Language Modeling with Pathways","code":"https://github.com/lucidrains/CoCa-pytorch","n_code_links":7,"syntology":{"n_ran":32,"n_unverified":5,"n_samples":37,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"ChatQA-1.5-llama3-8B (Zero-Shot, KILT)","metrics":{"EM":"81.0"},"uses_additional_data":true,"paper_date":"2024-01-18","paper":"/paper/chatqa-building-gpt-4-level-conversational-qa","paper_url":"https://arxiv.org/abs/2401.10225v5","paper_title":"ChatQA: Surpassing GPT-4 on Conversational QA and RAG","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":14,"model":"GaC(Qwen2-72B-Instruct + Llama-3-70B-Instruct)","metrics":{"EM":"79.29"},"uses_additional_data":false,"paper_date":"2024-06-18","paper":"/paper/breaking-the-ceiling-of-the-llm-community-by","paper_url":"https://arxiv.org/abs/2406.12585v2","paper_title":"Breaking the Ceiling of the LLM Community by Treating Token Generation as a Classification for Ensembling","code":"https://github.com/yaoching0/gac","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":3,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"Claude Instant 1.1 (few-shot, k=5)","metrics":{"EM":"78.9"},"uses_additional_data":false,"paper_date":"2023-07-11","paper":"/paper/model-card-and-evaluations-for-claude-models","paper_url":"https://www-files.anthropic.com/production/images/Model-Card-Claude-2.pdf","paper_title":"Model Card and Evaluations for Claude Models","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":16,"model":"code-davinci-002 175B + REPLUG LSR (Few-Shot)","metrics":{"EM":"77.3"},"uses_additional_data":false,"paper_date":"2023-01-30","paper":"/paper/replug-retrieval-augmented-black-box-language","paper_url":"https://arxiv.org/abs/2301.12652v4","paper_title":"REPLUG: Retrieval-Augmented Black-Box Language Models","code":"https://github.com/ruc-nlpir/flashrag","n_code_links":3,"syntology":{"n_ran":10,"n_unverified":3,"n_samples":13,"n_pointer_only_licence":13}},{"rank_in_archive_order":17,"model":"PaLM-540B (Zero-Shot)","metrics":{"EM":"76.9"},"uses_additional_data":false,"paper_date":"2022-04-05","paper":"/paper/palm-scaling-language-modeling-with-pathways-1","paper_url":"https://arxiv.org/abs/2204.02311v5","paper_title":"PaLM: Scaling Language Modeling with Pathways","code":"https://github.com/lucidrains/CoCa-pytorch","n_code_links":7,"syntology":{"n_ran":32,"n_unverified":5,"n_samples":37,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"code-davinci-002 175B + REPLUG (Few-Shot)","metrics":{"EM":"76.8"},"uses_additional_data":false,"paper_date":"2023-01-30","paper":"/paper/replug-retrieval-augmented-black-box-language","paper_url":"https://arxiv.org/abs/2301.12652v4","paper_title":"REPLUG: Retrieval-Augmented Black-Box Language Models","code":"https://github.com/ruc-nlpir/flashrag","n_code_links":3,"syntology":{"n_ran":10,"n_unverified":3,"n_samples":13,"n_pointer_only_licence":13}},{"rank_in_archive_order":19,"model":"GLaM 62B/64E (One-shot)","metrics":{"EM":"75.8"},"uses_additional_data":true,"paper_date":"2021-12-13","paper":"/paper/glam-efficient-scaling-of-language-models","paper_url":"https://arxiv.org/abs/2112.06905v2","paper_title":"GLaM: Efficient Scaling of Language Models with Mixture-of-Experts","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":20,"model":"GLaM 62B/64E (Few-shot)","metrics":{"EM":"75.8"},"uses_additional_data":false,"paper_date":"2021-12-13","paper":"/paper/glam-efficient-scaling-of-language-models","paper_url":"https://arxiv.org/abs/2112.06905v2","paper_title":"GLaM: Efficient Scaling of Language Models with Mixture-of-Experts","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":21,"model":"RA-DIT (Zero-Shot)","metrics":{"EM":"75.4"},"uses_additional_data":true,"paper_date":"2023-10-02","paper":"/paper/ra-dit-retrieval-augmented-dual-instruction","paper_url":"https://arxiv.org/abs/2310.01352v4","paper_title":"RA-DIT: Retrieval-Augmented Dual Instruction Tuning","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":22,"model":"PaLM 2-S (one-shot)","metrics":{"EM":"75.2"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/palm-2-technical-report-1","paper_url":"https://arxiv.org/abs/2305.10403v3","paper_title":"PaLM 2 Technical Report","code":"https://github.com/eternityyw/tram-benchmark","n_code_links":1,"syntology":null},{"rank_in_archive_order":23,"model":"LLaMA 65B (few-shot, k=64)","metrics":{"EM":"73.0"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":24,"model":"FiE+PAQ","metrics":{"EM":"72.6"},"uses_additional_data":false,"paper_date":"2022-11-18","paper":"/paper/fie-building-a-global-probability-space-by","paper_url":"https://arxiv.org/abs/2211.10147v1","paper_title":"FiE: Building a Global Probability Space by Leveraging Early Fusion in Encoder for Open-Domain Question Answering","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":25,"model":"LLaMA 65B (few-shot, k=5)","metrics":{"EM":"72.6"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":26,"model":"RankRAG-llama3-70b (Zero-Shot, DPR)","metrics":{"EM":"72.6"},"uses_additional_data":true,"paper_date":"2024-07-02","paper":"/paper/rankrag-unifying-context-ranking-with","paper_url":"https://arxiv.org/abs/2407.02485v1","paper_title":"RankRAG: Unifying Context Ranking with Retrieval-Augmented Generation in LLMs","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":27,"model":"FiD+Distil","metrics":{"EM":"72.1"},"uses_additional_data":true,"paper_date":"2020-12-08","paper":"/paper/distilling-knowledge-from-reader-to-retriever-1","paper_url":"https://arxiv.org/abs/2012.04584v2","paper_title":"Distilling Knowledge from Reader to Retriever for Question Answering","code":"https://github.com/facebookresearch/FiD","n_code_links":4,"syntology":{"n_ran":1,"n_unverified":0,"n_samples":1,"n_pointer_only_licence":1}},{"rank_in_archive_order":28,"model":"LLaMA 65B (one-shot)","metrics":{"EM":"71.6"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":29,"model":"EMDR2","metrics":{"EM":"71.4"},"uses_additional_data":false,"paper_date":"2021-06-09","paper":"/paper/end-to-end-training-of-multi-document-reader","paper_url":"https://arxiv.org/abs/2106.05346v2","paper_title":"End-to-End Training of Multi-Document Reader and Retriever for Open-Domain Question Answering","code":"https://github.com/DevSinghSachan/emdr2","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":0,"n_samples":4,"n_pointer_only_licence":4}},{"rank_in_archive_order":30,"model":"GLaM 62B/64E (Zero-shot)","metrics":{"EM":"71.3"},"uses_additional_data":false,"paper_date":"2021-12-13","paper":"/paper/glam-efficient-scaling-of-language-models","paper_url":"https://arxiv.org/abs/2112.06905v2","paper_title":"GLaM: Efficient Scaling of Language Models with Mixture-of-Experts","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":31,"model":"GPT-3 175B (Few-Shot)","metrics":{"EM":"71.2"},"uses_additional_data":true,"paper_date":"2020-05-28","paper":"/paper/language-models-are-few-shot-learners","paper_url":"https://arxiv.org/abs/2005.14165v4","paper_title":"Language Models are Few-Shot Learners","code":"https://github.com/ggml-org/llama.cpp","n_code_links":67,"syntology":{"n_ran":41,"n_unverified":24,"n_samples":65,"n_pointer_only_licence":4}},{"rank_in_archive_order":32,"model":"Mistral 7B (5-shot)","metrics":{"EM":"69.9"},"uses_additional_data":false,"paper_date":"2023-10-10","paper":"/paper/mistral-7b","paper_url":"https://arxiv.org/abs/2310.06825v1","paper_title":"Mistral 7B","code":"https://github.com/mistralai/mistral-src","n_code_links":6,"syntology":{"n_ran":10,"n_unverified":1,"n_samples":11,"n_pointer_only_licence":1}},{"rank_in_archive_order":33,"model":"ChatQA-1.5-llama3-70b (Zero-Shot, DPR)","metrics":{"EM":"69.0"},"uses_additional_data":false,"paper_date":"2024-01-18","paper":"/paper/chatqa-building-gpt-4-level-conversational-qa","paper_url":"https://arxiv.org/abs/2401.10225v5","paper_title":"ChatQA: Surpassing GPT-4 on Conversational QA and RAG","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":34,"model":"LLaMA 65B (zero-shot)","metrics":{"EM":"68.2"},"uses_additional_data":false,"paper_date":"2023-02-27","paper":"/paper/llama-open-and-efficient-foundation-language-1","paper_url":"https://arxiv.org/abs/2302.13971v1","paper_title":"LLaMA: Open and Efficient Foundation Language Models","code":"https://github.com/huggingface/transformers","n_code_links":57,"syntology":{"n_ran":37,"n_unverified":21,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":35,"model":"Fusion-in-Decoder (large)","metrics":{"EM":"67.6"},"uses_additional_data":false,"paper_date":"2020-07-02","paper":"/paper/leveraging-passage-retrieval-with-generative","paper_url":"https://arxiv.org/abs/2007.01282v2","paper_title":"Leveraging Passage Retrieval with Generative Models for Open Domain Question Answering","code":"https://github.com/jhyuklee/DensePhrases","n_code_links":8,"syntology":null},{"rank_in_archive_order":36,"model":"MemoReader","metrics":{"EM":"67.21","F1":"73.26"},"uses_additional_data":true,"paper_date":"2018-10-01","paper":"/paper/memoreader-large-scale-reading-comprehension","paper_url":"https://aclanthology.org/D18-1237","paper_title":"MemoReader: Large-Scale Reading Comprehension through Neural Memory Controller","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":37,"model":"S-Norm","metrics":{"EM":"66.37","F1":"71.32"},"uses_additional_data":true,"paper_date":"2017-10-29","paper":"/paper/simple-and-effective-multi-paragraph-reading","paper_url":"http://arxiv.org/abs/1710.10723v2","paper_title":"Simple and Effective Multi-Paragraph Reading Comprehension","code":"https://github.com/allenai/document-qa","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":4,"n_samples":4,"n_pointer_only_licence":0}},{"rank_in_archive_order":38,"model":"TOME-2","metrics":{"EM":"65.8"},"uses_additional_data":false,"paper_date":"2021-10-12","paper":"/paper/mention-memory-incorporating-textual-1","paper_url":"https://arxiv.org/abs/2110.06176v2","paper_title":"Mention Memory: incorporating textual knowledge into Transformers through entity mention attention","code":"https://github.com/google-research/language/tree/master/language/mentionmemory","n_code_links":1,"syntology":null},{"rank_in_archive_order":39,"model":"Shakti-LLM (2.5B)","metrics":{"EM":"58.2"},"uses_additional_data":false,"paper_date":"2024-10-15","paper":"/paper/shakti-a-2-5-billion-parameter-small-language","paper_url":"https://arxiv.org/abs/2410.11331v1","paper_title":"SHAKTI: A 2.5 Billion Parameter Small Language Model Optimized for Edge AI and Low-Resource Environments","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":40,"model":"Branch-Train-MiX 4x7B (sampling top-2 experts)","metrics":{"EM":"57.1"},"uses_additional_data":false,"paper_date":"2024-03-12","paper":"/paper/branch-train-mix-mixing-expert-llms-into-a","paper_url":"https://arxiv.org/abs/2403.07816v1","paper_title":"Branch-Train-MiX: Mixing Expert LLMs into a Mixture-of-Experts LLM","code":"https://github.com/Leeroo-AI/mergoo","n_code_links":1,"syntology":null},{"rank_in_archive_order":41,"model":"DPR","metrics":{"EM":"56.8"},"uses_additional_data":false,"paper_date":"2020-04-10","paper":"/paper/dense-passage-retrieval-for-open-domain","paper_url":"https://arxiv.org/abs/2004.04906v3","paper_title":"Dense Passage Retrieval for Open-Domain Question Answering","code":"https://github.com/huggingface/transformers","n_code_links":19,"syntology":{"n_ran":11,"n_unverified":3,"n_samples":14,"n_pointer_only_licence":9}},{"rank_in_archive_order":42,"model":"FLAN 137B (zero-shot)","metrics":{"EM":"56.7"},"uses_additional_data":false,"paper_date":"2021-09-03","paper":"/paper/finetuned-language-models-are-zero-shot","paper_url":"https://arxiv.org/abs/2109.01652v5","paper_title":"Finetuned Language Models Are Zero-Shot Learners","code":"https://github.com/hiyouga/llama-efficient-tuning","n_code_links":8,"syntology":{"n_ran":0,"n_unverified":1,"n_samples":1,"n_pointer_only_licence":0}},{"rank_in_archive_order":43,"model":"RAG","metrics":{"EM":"56.1"},"uses_additional_data":false,"paper_date":"2020-05-22","paper":"/paper/retrieval-augmented-generation-for-knowledge","paper_url":"https://arxiv.org/abs/2005.11401v4","paper_title":"Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks","code":"https://github.com/huggingface/transformers","n_code_links":18,"syntology":{"n_ran":6,"n_unverified":0,"n_samples":6,"n_pointer_only_licence":0}},{"rank_in_archive_order":44,"model":"Reading Twice for NLU","metrics":{"EM":"50.56","F1":"56.73"},"uses_additional_data":false,"paper_date":"2017-06-08","paper":"/paper/dynamic-integration-of-background-knowledge","paper_url":"http://arxiv.org/abs/1706.02596v3","paper_title":"Dynamic Integration of Background Knowledge in Neural NLU Systems","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":45,"model":"Mnemonic Reader","metrics":{"EM":"46.94","F1":"52.85"},"uses_additional_data":false,"paper_date":"2017-05-08","paper":"/paper/reinforced-mnemonic-reader-for-machine","paper_url":"http://arxiv.org/abs/1705.02798v6","paper_title":"Reinforced Mnemonic Reader for Machine Reading Comprehension","code":"https://github.com/HKUST-KnowComp/MnemonicReader","n_code_links":3,"syntology":null},{"rank_in_archive_order":46,"model":"ORQA","metrics":{"EM":"45"},"uses_additional_data":false,"paper_date":"2019-06-01","paper":"/paper/190600300","paper_url":"https://arxiv.org/abs/1906.00300v3","paper_title":"Latent Retrieval for Weakly Supervised Open Domain Question Answering","code":"https://github.com/google-research/language/tree/master/language/orqa","n_code_links":3,"syntology":null},{"rank_in_archive_order":47,"model":"MEMEN","metrics":{"EM":"43.16","F1":"46.90"},"uses_additional_data":false,"paper_date":"2017-07-28","paper":"/paper/memen-multi-layer-embedding-with-memory","paper_url":"http://arxiv.org/abs/1707.09098v1","paper_title":"MEMEN: Multi-layer Embedding with Memory Networks for Machine Comprehension","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":48,"model":"SpanBERT","metrics":{"F1":"83.6"},"uses_additional_data":false,"paper_date":"2019-07-24","paper":"/paper/spanbert-improving-pre-training-by","paper_url":"https://arxiv.org/abs/1907.10529v3","paper_title":"SpanBERT: Improving Pre-training by Representing and Predicting Spans","code":"https://github.com/facebookresearch/SpanBERT","n_code_links":6,"syntology":{"n_ran":6,"n_unverified":9,"n_samples":15,"n_pointer_only_licence":4}},{"rank_in_archive_order":49,"model":"BigBird-etc","metrics":{"F1":"80.9"},"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":50,"model":"DPA-RAG","metrics":{"F1":"80.1"},"uses_additional_data":false,"paper_date":"2024-06-26","paper":"/paper/understand-what-llm-needs-dual-preference","paper_url":"https://arxiv.org/abs/2406.18676v2","paper_title":"Understand What LLM Needs: Dual Preference Alignment for Retrieval-Augmented Generation","code":"https://github.com/dongguanting/dpa-rag","n_code_links":1,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":51,"model":"LinkBERT (large)","metrics":{"F1":"78.2"},"uses_additional_data":false,"paper_date":"2022-03-29","paper":"/paper/linkbert-pretraining-language-models-with","paper_url":"https://arxiv.org/abs/2203.15827v1","paper_title":"LinkBERT: Pretraining Language Models with Document Links","code":"https://github.com/michiyasunaga/LinkBERT","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":11,"n_samples":14,"n_pointer_only_licence":0}},{"rank_in_archive_order":52,"model":"DyREX","metrics":{"F1":"77.37"},"uses_additional_data":false,"paper_date":"2022-10-26","paper":"/paper/dyrex-dynamic-query-representation-for","paper_url":"https://arxiv.org/abs/2210.15048v1","paper_title":"DyREx: Dynamic Query Representation for Extractive Question Answering","code":"https://github.com/urchade/dyrex","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":53,"model":"Search-o1","metrics":{"F1":"74.1"},"uses_additional_data":false,"paper_date":"2025-01-09","paper":"/paper/search-o1-agentic-search-enhanced-large","paper_url":"https://arxiv.org/abs/2501.05366v1","paper_title":"Search-o1: Agentic Search-Enhanced Large Reasoning Models","code":"https://github.com/sunnynexus/search-o1","n_code_links":2,"syntology":null},{"rank_in_archive_order":54,"model":"UnitedQA (Hybrid reader)","metrics":{"F1":"70.3"},"uses_additional_data":false,"paper_date":"2021-01-01","paper":"/paper/unitedqa-a-hybrid-approach-for-open-domain","paper_url":"https://arxiv.org/abs/2101.00178v2","paper_title":"UnitedQA: A Hybrid Approach for Open Domain Question Answering","code":null,"n_code_links":0,"syntology":null},{"rank_in_archive_order":55,"model":"ReasonBERTR","metrics":{"F1":"45.5"},"uses_additional_data":false,"paper_date":"2021-09-10","paper":"/paper/reasonbert-pre-trained-to-reason-with-distant","paper_url":"https://arxiv.org/abs/2109.04912v1","paper_title":"ReasonBERT: Pre-trained to Reason with Distant Supervision","code":"https://github.com/sunlab-osu/reasonbert","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":8,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":56,"model":"ReasonBERTB","metrics":{"F1":"37.2"},"uses_additional_data":false,"paper_date":"2021-09-10","paper":"/paper/reasonbert-pre-trained-to-reason-with-distant","paper_url":"https://arxiv.org/abs/2109.04912v1","paper_title":"ReasonBERT: Pre-trained to Reason with Distant Supervision","code":"https://github.com/sunlab-osu/reasonbert","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":8,"n_samples":11,"n_pointer_only_licence":0}}],"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":27,"rows_with_any_sample_ran":25,"distinct_papers_with_graph_line":20,"distinct_papers_with_any_sample_ran":18,"samples_over_distinct_papers":{"n_ran":217,"n_unverified":122,"n_samples":339,"n_pointer_only_licence":74,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":405,"n_unverified":206,"n_samples":611,"n_pointer_only_licence":99,"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"}}}