{"url":"/sota/question-answering-on-social-iqa","task":{"name":"Question Answering","url":"/task/question-answering","note":null},"dataset":{"name":"SIQA","url":"/dataset/social-iqa"},"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":["Accuracy"],"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"}},"counts":{"rows":24,"rows_with_code":24,"rows_with_paper_page":24,"rows_dated":24,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"Unicorn 11B (fine-tuned)","metrics":{"Accuracy":"83.2"},"uses_additional_data":false,"paper_date":"2021-03-24","paper":"/paper/unicorn-on-rainbow-a-universal-commonsense","paper_url":"https://arxiv.org/abs/2103.13009v1","paper_title":"UNICORN on RAINBOW: A Universal Commonsense Reasoning Model on a New Multitask Benchmark","code":"https://github.com/allenai/rainbow","n_code_links":1,"syntology":{"n_ran":3,"n_unverified":2,"n_samples":5,"n_pointer_only_licence":0}},{"rank_in_archive_order":2,"model":"LLaMA-2 13B + MixLoRA","metrics":{"Accuracy":"82.5"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":3,"model":"CompassMTL 567M with Tailor","metrics":{"Accuracy":"82.2"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":4,"model":"CompassMTL 567M","metrics":{"Accuracy":"81.7"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":5,"model":"LLaMA-3 8B+MoSLoRA (fine-tuned)","metrics":{"Accuracy":"81.0"},"uses_additional_data":false,"paper_date":"2024-06-16","paper":"/paper/mixture-of-subspaces-in-low-rank-adaptation","paper_url":"https://arxiv.org/abs/2406.11909v3","paper_title":"Mixture-of-Subspaces in Low-Rank Adaptation","code":"https://github.com/wutaiqiang/moslora","n_code_links":1,"syntology":{"n_ran":4,"n_unverified":2,"n_samples":6,"n_pointer_only_licence":6}},{"rank_in_archive_order":6,"model":"DeBERTa-Large 304M","metrics":{"Accuracy":"80.2"},"uses_additional_data":false,"paper_date":"2022-10-29","paper":"/paper/two-is-better-than-many-binary-classification","paper_url":"https://arxiv.org/abs/2210.16495v1","paper_title":"Two is Better than Many? Binary Classification as an Effective Approach to Multi-Choice Question Answering","code":"https://github.com/declare-lab/team","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":7,"model":"DeBERTa-Large 304M (classification-based)","metrics":{"Accuracy":"79.9"},"uses_additional_data":false,"paper_date":"2022-10-29","paper":"/paper/two-is-better-than-many-binary-classification","paper_url":"https://arxiv.org/abs/2210.16495v1","paper_title":"Two is Better than Many? Binary Classification as an Effective Approach to Multi-Choice Question Answering","code":"https://github.com/declare-lab/team","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":5,"n_samples":7,"n_pointer_only_licence":0}},{"rank_in_archive_order":8,"model":"UnifiedQA 3B","metrics":{"Accuracy":"79.8"},"uses_additional_data":false,"paper_date":"2020-05-02","paper":"/paper/unifiedqa-crossing-format-boundaries-with-a","paper_url":"https://arxiv.org/abs/2005.00700v3","paper_title":"UnifiedQA: Crossing Format Boundaries With a Single QA System","code":"https://github.com/allenai/unifiedqa","n_code_links":2,"syntology":{"n_ran":4,"n_unverified":3,"n_samples":7,"n_pointer_only_licence":1}},{"rank_in_archive_order":9,"model":"ExDeBERTa 567M","metrics":{"Accuracy":"79.6"},"uses_additional_data":false,"paper_date":"2022-10-12","paper":"/paper/task-compass-scaling-multi-task-pre-training","paper_url":"https://arxiv.org/abs/2210.06277v1","paper_title":"Task Compass: Scaling Multi-task Pre-training with Task Prefix","code":"https://github.com/cooelf/compassmtl","n_code_links":1,"syntology":null},{"rank_in_archive_order":10,"model":"LLaMA-3 8B + MixLoRA","metrics":{"Accuracy":"78.8"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"LLaMA-2 7B + MixLoRA","metrics":{"Accuracy":"78"},"uses_additional_data":false,"paper_date":"2024-04-22","paper":"/paper/mixlora-enhancing-large-language-models-fine","paper_url":"https://arxiv.org/abs/2404.15159v3","paper_title":"MixLoRA: Enhancing Large Language Models Fine-Tuning with LoRA-based Mixture of Experts","code":"https://github.com/TUDB-Labs/MixLoRA","n_code_links":2,"syntology":{"n_ran":6,"n_unverified":5,"n_samples":11,"n_pointer_only_licence":0}},{"rank_in_archive_order":12,"model":"RoBERTa-Large 355M (fine-tuned)","metrics":{"Accuracy":"76.7"},"uses_additional_data":false,"paper_date":"2019-07-26","paper":"/paper/roberta-a-robustly-optimized-bert-pretraining","paper_url":"https://arxiv.org/abs/1907.11692v1","paper_title":"RoBERTa: A Robustly Optimized BERT Pretraining Approach","code":"https://github.com/huggingface/transformers","n_code_links":67,"syntology":{"n_ran":37,"n_unverified":11,"n_samples":48,"n_pointer_only_licence":23}},{"rank_in_archive_order":13,"model":"BERT-large 340M (fine-tuned)","metrics":{"Accuracy":"64.5"},"uses_additional_data":false,"paper_date":"2019-04-22","paper":"/paper/socialiqa-commonsense-reasoning-about-social","paper_url":"https://arxiv.org/abs/1904.09728v3","paper_title":"SocialIQA: Commonsense Reasoning about Social Interactions","code":"https://github.com/clear-nus/llm-human-model","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":14,"model":"BERT-base 110M (fine-tuned)","metrics":{"Accuracy":"63.1"},"uses_additional_data":false,"paper_date":"2019-04-22","paper":"/paper/socialiqa-commonsense-reasoning-about-social","paper_url":"https://arxiv.org/abs/1904.09728v3","paper_title":"SocialIQA: Commonsense Reasoning about Social Interactions","code":"https://github.com/clear-nus/llm-human-model","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":15,"model":"GPT-1 117M (fine-tuned)","metrics":{"Accuracy":"63"},"uses_additional_data":false,"paper_date":"2019-04-22","paper":"/paper/socialiqa-commonsense-reasoning-about-social","paper_url":"https://arxiv.org/abs/1904.09728v3","paper_title":"SocialIQA: Commonsense Reasoning about Social Interactions","code":"https://github.com/clear-nus/llm-human-model","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"phi-1.5-web 1.3B (zero-shot)","metrics":{"Accuracy":"53.0"},"uses_additional_data":false,"paper_date":"2023-09-11","paper":"/paper/textbooks-are-all-you-need-ii-phi-1-5","paper_url":"https://arxiv.org/abs/2309.05463v1","paper_title":"Textbooks Are All You Need II: phi-1.5 technical report","code":"https://github.com/knowlab/bi-weekly-paper-presentation","n_code_links":1,"syntology":null},{"rank_in_archive_order":17,"model":"phi-1.5 1.3B (zero-shot)","metrics":{"Accuracy":"52.6"},"uses_additional_data":false,"paper_date":"2023-09-11","paper":"/paper/textbooks-are-all-you-need-ii-phi-1-5","paper_url":"https://arxiv.org/abs/2309.05463v1","paper_title":"Textbooks Are All You Need II: phi-1.5 technical report","code":"https://github.com/knowlab/bi-weekly-paper-presentation","n_code_links":1,"syntology":null},{"rank_in_archive_order":18,"model":"LLaMA 65B (zero-shot)","metrics":{"Accuracy":"52.3"},"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":19,"model":"Chinchilla (zero-shot)","metrics":{"Accuracy":"51.3"},"uses_additional_data":false,"paper_date":"2022-03-29","paper":"/paper/training-compute-optimal-large-language","paper_url":"https://arxiv.org/abs/2203.15556v1","paper_title":"Training Compute-Optimal Large Language Models","code":"https://github.com/karpathy/llama2.c","n_code_links":2,"syntology":{"n_ran":8,"n_unverified":3,"n_samples":11,"n_pointer_only_licence":4}},{"rank_in_archive_order":20,"model":"Gopher (zero-shot)","metrics":{"Accuracy":"50.6"},"uses_additional_data":false,"paper_date":"2021-12-08","paper":"/paper/scaling-language-models-methods-analysis-1","paper_url":"https://arxiv.org/abs/2112.11446v2","paper_title":"Scaling Language Models: Methods, Analysis & Insights from Training Gopher","code":"https://github.com/allenai/dolma","n_code_links":3,"syntology":null},{"rank_in_archive_order":21,"model":"LLaMA 13B (zero-shot)","metrics":{"Accuracy":"50.4"},"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":22,"model":"LLaMA 33B (zero-shot)","metrics":{"Accuracy":"50.4"},"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":23,"model":"LLaMA 7B (zero-shot)","metrics":{"Accuracy":"48.9"},"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":"Random chance baseline","metrics":{"Accuracy":"33.3"},"uses_additional_data":false,"paper_date":"2019-04-22","paper":"/paper/socialiqa-commonsense-reasoning-about-social","paper_url":"https://arxiv.org/abs/1904.09728v3","paper_title":"SocialIQA: Commonsense Reasoning about Social Interactions","code":"https://github.com/clear-nus/llm-human-model","n_code_links":1,"syntology":{"n_ran":5,"n_unverified":5,"n_samples":10,"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":18,"rows_with_any_sample_ran":18,"distinct_papers_with_graph_line":9,"distinct_papers_with_any_sample_ran":9,"samples_over_distinct_papers":{"n_ran":106,"n_unverified":57,"n_samples":163,"n_pointer_only_licence":38,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":246,"n_unverified":150,"n_samples":396,"n_pointer_only_licence":50,"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"}}}