{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:BE2C67BZ22NTWFRDUWMEDCM4YR","short_pith_number":"pith:BE2C67BZ","schema_version":"1.0","canonical_sha256":"09342f7c39d69b3b1623a59841899cc4768991cf300207a8bfff3df975c1f13f","source":{"kind":"arxiv","id":"2010.11918","version":2},"attestation_state":"computed","paper":{"title":"AdapterDrop: On the Efficiency of Adapters in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andreas R\\\"uckl\\'e, Gregor Geigle, Iryna Gurevych, Jonas Pfeiffer, Max Glockner, Nils Reimers, Tilman Beck","submitted_at":"2020-10-22T17:49:42Z","abstract_excerpt":"Massively pre-trained transformer models are computationally expensive to fine-tune, slow for inference, and have large storage requirements. Recent approaches tackle these shortcomings by training smaller models, dynamically reducing the model size, and by training light-weight adapters. In this paper, we propose AdapterDrop, removing adapters from lower transformer layers during training and inference, which incorporates concepts from all three directions. We show that AdapterDrop can dynamically reduce the computational overhead when performing inference over multiple tasks simultaneously, "},"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":"2010.11918","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2020-10-22T17:49:42Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"b79db233884ad93f7085d12b2591a4a910db10e6d8ebc5763899dc6cd28298ab","abstract_canon_sha256":"ef80a276b74e36aeb2a26cd74f366d6d9ff4a856266e222340152ba56c0d8381"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:20:18.981402Z","signature_b64":"/dnUZrQDN/XK6BULg3PeXOyVD4wOgJl5dRvvoNlgEO/4dZwsa1nY8mOfvrsGd0Z2B1AZlSHrwSSdrm1wjUDNCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"09342f7c39d69b3b1623a59841899cc4768991cf300207a8bfff3df975c1f13f","last_reissued_at":"2026-07-05T03:20:18.980986Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:20:18.980986Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdapterDrop: On the Efficiency of Adapters in Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Andreas R\\\"uckl\\'e, Gregor Geigle, Iryna Gurevych, Jonas Pfeiffer, Max Glockner, Nils Reimers, Tilman Beck","submitted_at":"2020-10-22T17:49:42Z","abstract_excerpt":"Massively pre-trained transformer models are computationally expensive to fine-tune, slow for inference, and have large storage requirements. Recent approaches tackle these shortcomings by training smaller models, dynamically reducing the model size, and by training light-weight adapters. In this paper, we propose AdapterDrop, removing adapters from lower transformer layers during training and inference, which incorporates concepts from all three directions. We show that AdapterDrop can dynamically reduce the computational overhead when performing inference over multiple tasks simultaneously, "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2010.11918","kind":"arxiv","version":2},"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/2010.11918/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":"2010.11918","created_at":"2026-07-05T03:20:18.981038+00:00"},{"alias_kind":"arxiv_version","alias_value":"2010.11918v2","created_at":"2026-07-05T03:20:18.981038+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2010.11918","created_at":"2026-07-05T03:20:18.981038+00:00"},{"alias_kind":"pith_short_12","alias_value":"BE2C67BZ22NT","created_at":"2026-07-05T03:20:18.981038+00:00"},{"alias_kind":"pith_short_16","alias_value":"BE2C67BZ22NTWFRD","created_at":"2026-07-05T03:20:18.981038+00:00"},{"alias_kind":"pith_short_8","alias_value":"BE2C67BZ","created_at":"2026-07-05T03:20:18.981038+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2403.14608","citing_title":"Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey","ref_index":117,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18124","citing_title":"TLoRA: Task-aware Low Rank Adaptation of Large Language Models","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR","json":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR.json","graph_json":"https://pith.science/api/pith-number/BE2C67BZ22NTWFRDUWMEDCM4YR/graph.json","events_json":"https://pith.science/api/pith-number/BE2C67BZ22NTWFRDUWMEDCM4YR/events.json","paper":"https://pith.science/paper/BE2C67BZ"},"agent_actions":{"view_html":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR","download_json":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR.json","view_paper":"https://pith.science/paper/BE2C67BZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2010.11918&json=true","fetch_graph":"https://pith.science/api/pith-number/BE2C67BZ22NTWFRDUWMEDCM4YR/graph.json","fetch_events":"https://pith.science/api/pith-number/BE2C67BZ22NTWFRDUWMEDCM4YR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR/action/storage_attestation","attest_author":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR/action/author_attestation","sign_citation":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR/action/citation_signature","submit_replication":"https://pith.science/pith/BE2C67BZ22NTWFRDUWMEDCM4YR/action/replication_record"}},"created_at":"2026-07-05T03:20:18.981038+00:00","updated_at":"2026-07-05T03:20:18.981038+00:00"}