{"url":"/sota/question-answering-on-truthfulqa","task":{"name":"Question Answering","url":"/task/question-answering","note":null},"dataset":{"name":"TruthfulQA","url":"/dataset/truthfulqa"},"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":["MC1","MC2","% true","% info","% true (GPT-judge)","BLEURT","ROUGE","BLEU","EM","Accuracy"],"metric_direction":{"note":"inferred from the metric name only (the archive records no direction); null = not inferred, chart draws points only","by_metric":{"MC1":null,"MC2":null,"% true":null,"% info":null,"% true (GPT-judge)":null,"BLEURT":null,"ROUGE":"higher","BLEU":"higher","EM":null,"Accuracy":"higher"}},"counts":{"rows":33,"rows_with_code":29,"rows_with_paper_page":30,"rows_dated":30,"rows_using_additional_data":0},"rows":[{"rank_in_archive_order":1,"model":"GPT-4 (RLHF)","metrics":{"MC1":"0.59"},"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":2,"n_unverified":3,"n_samples":5,"n_pointer_only_licence":1}},{"rank_in_archive_order":2,"model":"Mistral-7B-Instruct-v0.2 + TruthX","metrics":{"MC1":"0.56","MC2":"0.75"},"uses_additional_data":false,"paper_date":"2024-02-27","paper":"/paper/truthx-alleviating-hallucinations-by-editing","paper_url":"https://arxiv.org/abs/2402.17811v2","paper_title":"TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space","code":"https://github.com/ictnlp/truthx","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":3,"model":"LLaMa-2-7B-Chat + TruthX","metrics":{"MC1":"0.54","MC2":"0.74"},"uses_additional_data":false,"paper_date":"2024-02-27","paper":"/paper/truthx-alleviating-hallucinations-by-editing","paper_url":"https://arxiv.org/abs/2402.17811v2","paper_title":"TruthX: Alleviating Hallucinations by Editing Large Language Models in Truthful Space","code":"https://github.com/ictnlp/truthx","n_code_links":1,"syntology":{"n_ran":2,"n_unverified":1,"n_samples":3,"n_pointer_only_licence":3}},{"rank_in_archive_order":4,"model":"LLaMA-2-Chat-13B + Representation Control (Contrast Vector)","metrics":{"MC1":"0.54"},"uses_additional_data":false,"paper_date":"2023-10-02","paper":"/paper/representation-engineering-a-top-down","paper_url":"https://arxiv.org/abs/2310.01405v4","paper_title":"Representation Engineering: A Top-Down Approach to AI Transparency","code":"https://github.com/andyzoujm/representation-engineering","n_code_links":5,"syntology":{"n_ran":4,"n_unverified":6,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":5,"model":"LLaMA-2-Chat-7B + Representation Control (Contrast Vector)","metrics":{"MC1":"0.48"},"uses_additional_data":false,"paper_date":"2023-10-02","paper":"/paper/representation-engineering-a-top-down","paper_url":"https://arxiv.org/abs/2310.01405v4","paper_title":"Representation Engineering: A Top-Down Approach to AI Transparency","code":"https://github.com/andyzoujm/representation-engineering","n_code_links":5,"syntology":{"n_ran":4,"n_unverified":6,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":6,"model":"Vicuna 7B + Inference Time Intervention (ITI)","metrics":{"% info":"83.5","% true":"88.6","MC1":"0.389"},"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":"Alpaca 7B + Inference Time Intervention (ITI)","metrics":{"% info":"97.7","% true":"66.6","MC1":"0.319"},"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":"Gopher 280B (zero-shot, Our Prompt + Choices)","metrics":{"MC1":"0.295"},"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":9,"model":"LLaMA 7B + Inference Time Intervention (ITI)","metrics":{"% info":"93.8","% true":"45.1","MC1":"0.288"},"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":"GAL 120B","metrics":{"MC1":"0.26"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":11,"model":"Gopher 7.1 (zero-shot, QA prompts)","metrics":{"MC1":"0.25"},"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":12,"model":"GAL 30B","metrics":{"MC1":"0.24"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":13,"model":"Gopher 7.1B (zero-shot, Our Prompt + Choices)","metrics":{"MC1":"0.23"},"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":14,"model":"Gopher 1.4 (zero-shot, QA prompts)","metrics":{"MC1":"0.23"},"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":15,"model":"GPT-2 1.5B","metrics":{"% info":"89.84","% true":"29.50","% true (GPT-judge)":"29.87","BLEU":"-4.91","BLEURT":"-0.25","MC1":"0.22","MC2":"0.39","ROUGE":"-9.41"},"uses_additional_data":false,"paper_date":"2021-09-08","paper":"/paper/truthfulqa-measuring-how-models-mimic-human","paper_url":"https://arxiv.org/abs/2109.07958v2","paper_title":"TruthfulQA: Measuring How Models Mimic Human Falsehoods","code":"https://github.com/sylinrl/truthfulqa","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":16,"model":"Gopher 1.4B (zero-shot, Our Prompt + Choices)","metrics":{"MC1":"0.217"},"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":17,"model":"GPT-3 175B","metrics":{"% info":"97.55","% true":"20.44","% true (GPT-judge)":"20.56","BLEU":"-17.38","BLEURT":"-0.56","MC1":"0.21","MC2":"0.33","ROUGE":"-17.75"},"uses_additional_data":false,"paper_date":"2021-09-08","paper":"/paper/truthfulqa-measuring-how-models-mimic-human","paper_url":"https://arxiv.org/abs/2109.07958v2","paper_title":"TruthfulQA: Measuring How Models Mimic Human Falsehoods","code":"https://github.com/sylinrl/truthfulqa","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":18,"model":"OPT 175B","metrics":{"MC1":"0.21"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":19,"model":"GPT-J 6B","metrics":{"% info":"89.96","% true":"26.68","% true (GPT-judge)":"27.17","BLEU":"-7.58","BLEURT":"-0.31","MC1":"0.20","MC2":"0.36","ROUGE":"-11.35"},"uses_additional_data":false,"paper_date":"2021-09-08","paper":"/paper/truthfulqa-measuring-how-models-mimic-human","paper_url":"https://arxiv.org/abs/2109.07958v2","paper_title":"TruthfulQA: Measuring How Models Mimic Human Falsehoods","code":"https://github.com/sylinrl/truthfulqa","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":20,"model":"UnifiedQA 3B","metrics":{"% info":"64.50","% true":"53.86","% true (GPT-judge)":"53.24","BLEU":"-0.16","BLEURT":"0.08","MC1":"0.19","MC2":"0.35","ROUGE":"1.76"},"uses_additional_data":false,"paper_date":"2021-09-08","paper":"/paper/truthfulqa-measuring-how-models-mimic-human","paper_url":"https://arxiv.org/abs/2109.07958v2","paper_title":"TruthfulQA: Measuring How Models Mimic Human Falsehoods","code":"https://github.com/sylinrl/truthfulqa","n_code_links":3,"syntology":{"n_ran":2,"n_unverified":0,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":21,"model":"GAL 125M","metrics":{"MC1":"0.19"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":22,"model":"GAL 1.3B","metrics":{"MC1":"0.19"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":23,"model":"GAL 6.7B","metrics":{"MC1":"0.19"},"uses_additional_data":false,"paper_date":"2022-11-16","paper":"/paper/galactica-a-large-language-model-for-science-1","paper_url":"https://arxiv.org/abs/2211.09085v1","paper_title":"Galactica: A Large Language Model for Science","code":"https://github.com/paperswithcode/galai","n_code_links":1,"syntology":{"n_ran":0,"n_unverified":2,"n_samples":2,"n_pointer_only_licence":0}},{"rank_in_archive_order":24,"model":"Gopher 280B (zero-shot, QA prompts)","metrics":{"MC1":"0. 27"},"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":25,"model":"LLaMA 65B","metrics":{"% info":"53","% true":"57"},"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":26,"n_unverified":32,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":26,"model":"LLaMA 33B","metrics":{"% info":"48","% true":"52"},"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":26,"n_unverified":32,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":27,"model":"LLaMA 13B","metrics":{"% info":"41","% true":"47"},"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":26,"n_unverified":32,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":28,"model":"LLaMA 7B","metrics":{"% info":"29","% true":"33"},"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":26,"n_unverified":32,"n_samples":58,"n_pointer_only_licence":4}},{"rank_in_archive_order":29,"model":"CoA","metrics":{"EM":"67.3"},"uses_additional_data":false,"paper_date":"2024-03-26","paper":"/paper/chain-of-action-faithful-and-multimodal","paper_url":"https://arxiv.org/abs/2403.17359v1","paper_title":"Chain-of-Action: Faithful and Multimodal Question Answering through Large Language Models","code":"https://github.com/MAGICS-LAB/Chain-of-Actions","n_code_links":1,"syntology":{"n_ran":8,"n_unverified":2,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":30,"model":"ToT","metrics":{"EM":"66.6"},"uses_additional_data":false,"paper_date":"2023-05-17","paper":"/paper/tree-of-thoughts-deliberate-problem-solving-1","paper_url":"https://arxiv.org/abs/2305.10601v2","paper_title":"Tree of Thoughts: Deliberate Problem Solving with Large Language Models","code":"https://github.com/ysymyth/tree-of-thought-llm","n_code_links":6,"syntology":{"n_ran":7,"n_unverified":17,"n_samples":24,"n_pointer_only_licence":0}},{"rank_in_archive_order":31,"model":"CoA w/o actions","metrics":{"EM":"63.3"},"uses_additional_data":false,"paper_date":"2024-03-26","paper":"/paper/chain-of-action-faithful-and-multimodal","paper_url":"https://arxiv.org/abs/2403.17359v1","paper_title":"Chain-of-Action: Faithful and Multimodal Question Answering through Large Language Models","code":"https://github.com/MAGICS-LAB/Chain-of-Actions","n_code_links":1,"syntology":{"n_ran":8,"n_unverified":2,"n_samples":10,"n_pointer_only_licence":0}},{"rank_in_archive_order":32,"model":"Auto-CoT","metrics":{"EM":"42.2"},"uses_additional_data":false,"paper_date":"2022-10-07","paper":"/paper/automatic-chain-of-thought-prompting-in-large","paper_url":"https://arxiv.org/abs/2210.03493v1","paper_title":"Automatic Chain of Thought Prompting in Large Language Models","code":"https://github.com/microsoft/guidance","n_code_links":5,"syntology":{"n_ran":1,"n_unverified":2,"n_samples":3,"n_pointer_only_licence":0}},{"rank_in_archive_order":33,"model":"Shakti-LLM (2.5B)","metrics":{"Accuracy":"68.4"},"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}],"since_archive":{"present":false,"note":"No Syntology-extracted rows are published in this build."},"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":23,"rows_with_any_sample_ran":17,"distinct_papers_with_graph_line":9,"distinct_papers_with_any_sample_ran":8,"samples_over_distinct_papers":{"n_ran":52,"n_unverified":65,"n_samples":117,"n_pointer_only_licence":8,"note":"each paper (arXiv id) counted once, however many rows it is behind; this is the page-level figure"},"samples_row_weighted":{"n_ran":150,"n_unverified":180,"n_samples":330,"n_pointer_only_licence":23,"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"}}}