{"url":"/dataset/peerqa","name":"PeerQA","full_name":null,"description_markdown":"We present PeerQA, a real-world, scientific, document-level Question Answering (QA) dataset. PeerQA questions have been sourced from peer reviews, which contain questions that reviewers raised while thoroughly examining the scientific article. Answers have been annotated by the original authors of each paper. The dataset contains 579 QA pairs from 208 academic articles, with a majority from ML and NLP, as well as a subset of other scientific communities like Geoscience and Public Health. PeerQA supports three critical tasks for developing practical QA systems: Evidence retrieval, unanswerable question classification, and answer generation. We provide a detailed analysis of the collected dataset and conduct experiments establishing baseline systems for all three tasks. Our experiments and analyses reveal the need for decontextualization in document-level retrieval, where we find that even simple decontextualization approaches consistently improve retrieval performance across architectures. On answer generation, PeerQA serves as a challenging benchmark for long-context modeling, as the papers have an average size of 12k tokens.","description_withheld":null,"homepage":"","introduced_date":"2025-02-19","introduced_date_note":null,"introduced_by":{"paper":"/paper/peerqa-a-scientific-question-answering","title":"PeerQA: A Scientific Question Answering Dataset from Peer Reviews","first_author":"Tim Baumgärtner","url":null},"license":{"name":"CC-BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/deed.en"},"modalities":[{"name":"Texts","url":"/datasets/modality/texts"}],"tasks":[{"name":"Question Answering","url":"/task/question-answering","datasets_with_task":"/datasets/task/question-answering"},{"name":"Passage Retrieval","url":"/task/passage-retrieval","datasets_with_task":"/datasets/task/passage-retrieval"},{"name":"Text Retrieval","url":"/task/text-retrieval","datasets_with_task":"/datasets/task/text-retrieval"},{"name":"Sentence Retrieval","url":"/task/sentence-retrieval","datasets_with_task":"/datasets/task/sentence-retrieval"},{"name":"answerability prediction","url":"/task/answerability-prediction","datasets_with_task":"/datasets/task/answerability-prediction"},{"name":"RAG","url":"/task/rag","datasets_with_task":"/datasets/task/rag"}],"languages":[{"name":"English","url":"/datasets/language/english"}],"variants":["PeerQA"],"data_loaders":[{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/UKPLab/HaluQuestQA","frameworks":["tf","pytorch","jax"]},{"repo":"https://github.com/huggingface/datasets","url":"https://huggingface.co/datasets/UKPLab/PeerQA","frameworks":["tf","pytorch","jax"]}],"num_papers_in_archive":12,"source":{"archive":"pwc-archive (Hugging Face), CC BY-SA 4.0","snapshot":"2025-07-28"},"benchmarks":[{"leaderboard":"/sota/passage-retrieval-on-peerqa","task":"Passage Retrieval","dataset_variant":"PeerQA","rows":8,"metrics":["Recall@10","MRR"],"first_row_in_archive_order":{"model":"SPLADEv3","paper":"/paper/splade-v3-new-baselines-for-splade","metrics":{"MRR":"0.4536","Recall@10":"0.6851"},"code_links":[{"title":"naver/splade","url":"https://github.com/naver/splade"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/answerability-prediction-on-peerqa","task":"answerability prediction","dataset_variant":"PeerQA","rows":6,"metrics":["Macro F1"],"first_row_in_archive_order":{"model":"Mistral-IT-v02-7B-32k","paper":"/paper/mistral-7b","metrics":{"Macro F1":"0.4703"},"code_links":[{"title":"mistralai/mistral-src","url":"https://github.com/mistralai/mistral-src"},{"title":"facebookresearch/fairseq2","url":"https://github.com/facebookresearch/fairseq2"},{"title":"mgmalek/efficient_cross_entropy","url":"https://github.com/mgmalek/efficient_cross_entropy"},{"title":"ninglab/ecellm","url":"https://github.com/ninglab/ecellm"},{"title":"knowlab/bi-weekly-paper-presentation","url":"https://github.com/knowlab/bi-weekly-paper-presentation"},{"title":"pwc-1/Paper-9","url":"https://github.com/pwc-1/Paper-9/tree/main/2/mistral"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"},{"leaderboard":"/sota/question-answering-on-peerqa","task":"Question Answering","dataset_variant":"PeerQA","rows":6,"metrics":["Prometheus-2 Answer Correctness","Rouge-L","AlignScore"],"first_row_in_archive_order":{"model":"GPT-4o-2024-08-06-128k","paper":"/paper/gpt-4-technical-report-1","metrics":{"AlignScore":"0.1224","Prometheus-2 Answer Correctness":"3.4612","Rouge-L":"0.2266"},"code_links":[{"title":"openai/evals","url":"https://github.com/openai/evals"},{"title":"shmsw25/factscore","url":"https://github.com/shmsw25/factscore"},{"title":"unispac/visual-adversarial-examples-jailbreak-large-language-models","url":"https://github.com/unispac/visual-adversarial-examples-jailbreak-large-language-models"},{"title":"gpt4life/alpagasus","url":"https://github.com/gpt4life/alpagasus"},{"title":"emrgnt-cmplxty/zero-shot-replication","url":"https://github.com/emrgnt-cmplxty/zero-shot-replication"},{"title":"ethz-privsec/superhuman-ai-consistency","url":"https://github.com/ethz-privsec/superhuman-ai-consistency"},{"title":"ethz-spylab/superhuman-ai-consistency","url":"https://github.com/ethz-spylab/superhuman-ai-consistency"},{"title":"eternityyw/tram-benchmark","url":"https://github.com/eternityyw/tram-benchmark"},{"title":"AUCOHL/RTL-Repo","url":"https://github.com/AUCOHL/RTL-Repo"},{"title":"zach-zhiling-zheng/reticular_chemist","url":"https://github.com/zach-zhiling-zheng/reticular_chemist"},{"title":"lflage/openfactscore","url":"https://github.com/lflage/openfactscore"}]},"note":"rows are the archive's own order at snapshot; nothing here re-ranks them"}],"papers_with_a_benchmark_row":[{"paper":"/paper/the-llama-3-herd-of-models","title":"The Llama 3 Herd of Models","date":"2024-07-31","rows_on_this_dataset":4,"code_links":5,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":9,"samples_ran":2,"samples_unverified":7,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/splade-v3-new-baselines-for-splade","title":"SPLADE-v3: New baselines for SPLADE","date":"2024-03-11","rows_on_this_dataset":1,"code_links":1,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":3,"samples_ran":3,"samples_unverified":0,"pointer_only_for_licence":3,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/mistral-7b","title":"Mistral 7B","date":"2023-10-10","rows_on_this_dataset":2,"code_links":6,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":11,"samples_ran":9,"samples_unverified":2,"pointer_only_for_licence":1,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/knowledge-distillation-of-large-language","title":"MiniLLM: Knowledge Distillation of Large Language Models","date":"2023-06-14","rows_on_this_dataset":1,"code_links":2,"syntology":null},{"paper":"/paper/gpt-4-technical-report-1","title":"GPT-4 Technical Report","date":"2023-03-15","rows_on_this_dataset":2,"code_links":11,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":5,"samples_ran":2,"samples_unverified":3,"pointer_only_for_licence":1,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/how-to-train-your-dragon-diverse-augmentation","title":"How to Train Your DRAGON: Diverse Augmentation Towards Generalizable Dense Retrieval","date":"2023-02-15","rows_on_this_dataset":1,"code_links":1,"syntology":null},{"paper":"/paper/towards-unsupervised-dense-information","title":"Unsupervised Dense Information Retrieval with Contrastive Learning","date":"2021-12-16","rows_on_this_dataset":2,"code_links":6,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":5,"samples_ran":1,"samples_unverified":4,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/large-dual-encoders-are-generalizable","title":"Large Dual Encoders Are Generalizable Retrievers","date":"2021-12-15","rows_on_this_dataset":1,"code_links":2,"syntology":null},{"paper":"/paper/colbertv2-effective-and-efficient-retrieval","title":"ColBERTv2: Effective and Efficient Retrieval via Lightweight Late Interaction","date":"2021-12-02","rows_on_this_dataset":1,"code_links":3,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":1,"samples_ran":1,"samples_unverified":0,"pointer_only_for_licence":0,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}},{"paper":"/paper/language-models-are-few-shot-learners","title":"Language Models are Few-Shot Learners","date":"2020-05-28","rows_on_this_dataset":2,"code_links":67,"syntology":{"read_at":"2026-09-24T18:15:14+00:00","samples_harvested":65,"samples_ran":15,"samples_unverified":50,"pointer_only_for_licence":4,"claim":"Per-sample execution on synthesized fixtures; not a correctness claim."}}],"syntology_totals":{"read_at":"2026-09-24T18:15:14+00:00","papers_with_samples":7,"samples_harvested":99,"samples_ran":33,"samples_unverified":66,"pointer_only_for_licence":9,"papers_with_no_sample_that_ran":0,"note":"the per-paper counts above, summed; not a rate"},"papers_note":"The archive never published its papers-using-dataset list; these are papers with a leaderboard row on this dataset's benchmarks."}