{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:QY4DZDTSM6OMXEIYYB3VGS3UVO","short_pith_number":"pith:QY4DZDTS","schema_version":"1.0","canonical_sha256":"86383c8e72679ccb9118c077534b74ab9feeb0b44fcf61a556acd54f938adeaf","source":{"kind":"arxiv","id":"2004.10964","version":3},"attestation_state":"computed","paper":{"title":"Don't Stop Pretraining: Adapt Language Models to Domains and Tasks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Ana Marasovi\\'c, Doug Downey, Iz Beltagy, Kyle Lo, Noah A. Smith, Suchin Gururangan, Swabha Swayamdipta","submitted_at":"2020-04-23T04:21:19Z","abstract_excerpt":"Language models pretrained on text from a wide variety of sources form the foundation of today's NLP. In light of the success of these broad-coverage models, we investigate whether it is still helpful to tailor a pretrained model to the domain of a target task. We present a study across four domains (biomedical and computer science publications, news, and reviews) and eight classification tasks, showing that a second phase of pretraining in-domain (domain-adaptive pretraining) leads to performance gains, under both high- and low-resource settings. Moreover, adapting to the task's unlabeled dat"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2004.10964","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-23T04:21:19Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"133bc908d3b0ee1a7b2a66710754704e5c34b899f4d8d0aef42f455efa52056a","abstract_canon_sha256":"b3b003bde521fe5ef2661cc59275c825232844a96f47fb3a431550b32521caa2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:00:51.014032Z","signature_b64":"Nu3KSWtLvxUndm2yjFXdBypbh1cUO+JaRv+GJkXOdAB+lQYURUEZ5iuxIJy/r37rR6Toj1ir8/uxV5U04Lr6BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"86383c8e72679ccb9118c077534b74ab9feeb0b44fcf61a556acd54f938adeaf","last_reissued_at":"2026-07-05T01:00:51.013634Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:00:51.013634Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Don't Stop Pretraining: Adapt Language Models to Domains and Tasks","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Ana Marasovi\\'c, Doug Downey, Iz Beltagy, Kyle Lo, Noah A. Smith, Suchin Gururangan, Swabha Swayamdipta","submitted_at":"2020-04-23T04:21:19Z","abstract_excerpt":"Language models pretrained on text from a wide variety of sources form the foundation of today's NLP. In light of the success of these broad-coverage models, we investigate whether it is still helpful to tailor a pretrained model to the domain of a target task. We present a study across four domains (biomedical and computer science publications, news, and reviews) and eight classification tasks, showing that a second phase of pretraining in-domain (domain-adaptive pretraining) leads to performance gains, under both high- and low-resource settings. Moreover, adapting to the task's unlabeled dat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.10964","kind":"arxiv","version":3},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2004.10964/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"2004.10964","created_at":"2026-07-05T01:00:51.013696+00:00"},{"alias_kind":"arxiv_version","alias_value":"2004.10964v3","created_at":"2026-07-05T01:00:51.013696+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.10964","created_at":"2026-07-05T01:00:51.013696+00:00"},{"alias_kind":"pith_short_12","alias_value":"QY4DZDTSM6OM","created_at":"2026-07-05T01:00:51.013696+00:00"},{"alias_kind":"pith_short_16","alias_value":"QY4DZDTSM6OMXEIY","created_at":"2026-07-05T01:00:51.013696+00:00"},{"alias_kind":"pith_short_8","alias_value":"QY4DZDTS","created_at":"2026-07-05T01:00:51.013696+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":21,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2606.07521","citing_title":"Evaluating Hallucinations in Domain-Adapted Large Language Models","ref_index":7,"is_internal_anchor":true},{"citing_arxiv_id":"2606.18852","citing_title":"Aligning Implied Statements for Implicit Hate Speech Generalizability with Context-Bounded Semi-hard Negative Mining","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02100","citing_title":"PortBERT: Navigating the Depths of Portuguese Language Models","ref_index":118,"is_internal_anchor":false},{"citing_arxiv_id":"2606.02606","citing_title":"ReLoRA: Knowledge-Reusing Adaptation for Fast Rollout of Evolving LLM Services","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.24718","citing_title":"The Tokenizer Tax Across 25 European Languages: Domain Invariance, Cross-Lingual Few-Shot Effects, and the Ukrainian Penalty","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26601","citing_title":"FTibSuite: A Comprehensive Resource Suite for Tibetan Vision-Language Modeling","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2306.00890","citing_title":"LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2406.11354","citing_title":"Preserving Knowledge in Large Language Model with Model-Agnostic Self-Decompression","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2409.06624","citing_title":"A Practice of Post-Training on Llama-3 70B with Optimal Selection of Additional Language Mixture Ratio","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2412.19363","citing_title":"Large Language Models for Market Research: A Data-augmentation Approach","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2605.21984","citing_title":"Echo: Learning from Experience Data via User-Driven Refinement","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13415","citing_title":"KIT-TIP-NLP at MultiPride: Continual Learning with Multilingual Foundation Model","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2308.08089","citing_title":"DragNUWA: Fine-grained Control in Video Generation by Integrating Text, Image, and Trajectory","ref_index":285,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17762","citing_title":"Surface-Form Neural Sparse Retrieval: Robust Fuzzy Matching for Industrial Music Search","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2510.17934","citing_title":"AtlasKV: Augmenting LLMs with Billion-Scale Knowledge Graphs in 20GB VRAM","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2309.16671","citing_title":"Demystifying CLIP Data","ref_index":58,"is_internal_anchor":false},{"citing_arxiv_id":"2605.13415","citing_title":"KIT-TIP-NLP at MultiPride: Continual Learning with Multilingual Foundation Model","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10590","citing_title":"Bridging Linguistic Gaps: Cross-Lingual Mapping in Pre-Training and Dataset for Enhanced Multilingual LLM Performance","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12018","citing_title":"LLMs Struggle with Abstract Meaning Comprehension More Than Expected","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07938","citing_title":"Prototype Guided Post-pretraining for Single-Cell Representation Learning","ref_index":72,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14815","citing_title":"Domain Fine-Tuning FinBERT on Finnish Histopathological Reports: Train-Time Signals and Downstream Correlations","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO","json":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO.json","graph_json":"https://pith.science/api/pith-number/QY4DZDTSM6OMXEIYYB3VGS3UVO/graph.json","events_json":"https://pith.science/api/pith-number/QY4DZDTSM6OMXEIYYB3VGS3UVO/events.json","paper":"https://pith.science/paper/QY4DZDTS"},"agent_actions":{"view_html":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO","download_json":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO.json","view_paper":"https://pith.science/paper/QY4DZDTS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2004.10964&json=true","fetch_graph":"https://pith.science/api/pith-number/QY4DZDTSM6OMXEIYYB3VGS3UVO/graph.json","fetch_events":"https://pith.science/api/pith-number/QY4DZDTSM6OMXEIYYB3VGS3UVO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO/action/storage_attestation","attest_author":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO/action/author_attestation","sign_citation":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO/action/citation_signature","submit_replication":"https://pith.science/pith/QY4DZDTSM6OMXEIYYB3VGS3UVO/action/replication_record"}},"created_at":"2026-07-05T01:00:51.013696+00:00","updated_at":"2026-07-05T01:00:51.013696+00:00"}