{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:MNNWJO3YLRMPWVA5EX2ADTZMPM","short_pith_number":"pith:MNNWJO3Y","schema_version":"1.0","canonical_sha256":"635b64bb785c58fb541d25f401cf2c7b27bee3346cf181ca51c98f17973019fb","source":{"kind":"arxiv","id":"2005.00247","version":3},"attestation_state":"computed","paper":{"title":"AdapterFusion: Non-Destructive Task Composition for Transfer Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aishwarya Kamath, Andreas R\\\"uckl\\'e, Iryna Gurevych, Jonas Pfeiffer, Kyunghyun Cho","submitted_at":"2020-05-01T07:03:42Z","abstract_excerpt":"Sequential fine-tuning and multi-task learning are methods aiming to incorporate knowledge from multiple tasks; however, they suffer from catastrophic forgetting and difficulties in dataset balancing. To address these shortcomings, we propose AdapterFusion, a new two stage learning algorithm that leverages knowledge from multiple tasks. First, in the knowledge extraction stage we learn task specific parameters called adapters, that encapsulate the task-specific information. We then combine the adapters in a separate knowledge composition step. We show that by separating the two stages, i.e., k"},"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":"2005.00247","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-05-01T07:03:42Z","cross_cats_sorted":[],"title_canon_sha256":"279bbd39113cd89ab11478010ad7f817bb8ec0da24e0589af13ce50cea53c8db","abstract_canon_sha256":"2c69db1deeb4f88e9e1bd03163773c56d0efe9609585c8178ea56d44e1ef2535"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:09:39.147704Z","signature_b64":"8aQWHYeDh/07wd9+tJ6BvkVAmqIv3oZjWEXGpiepGWW0DZUuutZ4XaD8ihbJ+6lYE+hqA+Lg1RCgRVFSg0VxDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"635b64bb785c58fb541d25f401cf2c7b27bee3346cf181ca51c98f17973019fb","last_reissued_at":"2026-07-05T02:09:39.147362Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:09:39.147362Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"AdapterFusion: Non-Destructive Task Composition for Transfer Learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aishwarya Kamath, Andreas R\\\"uckl\\'e, Iryna Gurevych, Jonas Pfeiffer, Kyunghyun Cho","submitted_at":"2020-05-01T07:03:42Z","abstract_excerpt":"Sequential fine-tuning and multi-task learning are methods aiming to incorporate knowledge from multiple tasks; however, they suffer from catastrophic forgetting and difficulties in dataset balancing. To address these shortcomings, we propose AdapterFusion, a new two stage learning algorithm that leverages knowledge from multiple tasks. First, in the knowledge extraction stage we learn task specific parameters called adapters, that encapsulate the task-specific information. We then combine the adapters in a separate knowledge composition step. We show that by separating the two stages, i.e., k"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2005.00247","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/2005.00247/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":"2005.00247","created_at":"2026-07-05T02:09:39.147418+00:00"},{"alias_kind":"arxiv_version","alias_value":"2005.00247v3","created_at":"2026-07-05T02:09:39.147418+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2005.00247","created_at":"2026-07-05T02:09:39.147418+00:00"},{"alias_kind":"pith_short_12","alias_value":"MNNWJO3YLRMP","created_at":"2026-07-05T02:09:39.147418+00:00"},{"alias_kind":"pith_short_16","alias_value":"MNNWJO3YLRMPWVA5","created_at":"2026-07-05T02:09:39.147418+00:00"},{"alias_kind":"pith_short_8","alias_value":"MNNWJO3Y","created_at":"2026-07-05T02:09:39.147418+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":11,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.11712","citing_title":"Substrate Asymmetry in User-Side Memory: A Diagnostic Framework","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2501.05465","citing_title":"Small Language Models (SLMs) Can Still Pack a Punch: A survey (updated 2026)","ref_index":101,"is_internal_anchor":false},{"citing_arxiv_id":"2605.14055","citing_title":"PEML: Parameter-efficient Multi-Task Learning with Optimized Continuous Prompts","ref_index":81,"is_internal_anchor":false},{"citing_arxiv_id":"2303.16199","citing_title":"LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention","ref_index":168,"is_internal_anchor":false},{"citing_arxiv_id":"2403.14608","citing_title":"Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey","ref_index":35,"is_internal_anchor":false},{"citing_arxiv_id":"2303.10512","citing_title":"AdaLoRA: Adaptive Budget Allocation for Parameter-Efficient Fine-Tuning","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27596","citing_title":"SECOS: Semantic Capture for Rigorous Classification in Open-World Semi-Supervised Learning","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2101.00190","citing_title":"Prefix-Tuning: Optimizing Continuous Prompts for Generation","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2604.05257","citing_title":"Extending Tabular Denoising Diffusion Probabilistic Models for Time-Series Data Generation","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18124","citing_title":"TLoRA: Task-aware Low Rank Adaptation of Large Language Models","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2604.22838","citing_title":"Neural Network Optimization Reimagined: Decoupled Techniques for Scratch and Fine-Tuning","ref_index":66,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM","json":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM.json","graph_json":"https://pith.science/api/pith-number/MNNWJO3YLRMPWVA5EX2ADTZMPM/graph.json","events_json":"https://pith.science/api/pith-number/MNNWJO3YLRMPWVA5EX2ADTZMPM/events.json","paper":"https://pith.science/paper/MNNWJO3Y"},"agent_actions":{"view_html":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM","download_json":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM.json","view_paper":"https://pith.science/paper/MNNWJO3Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2005.00247&json=true","fetch_graph":"https://pith.science/api/pith-number/MNNWJO3YLRMPWVA5EX2ADTZMPM/graph.json","fetch_events":"https://pith.science/api/pith-number/MNNWJO3YLRMPWVA5EX2ADTZMPM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM/action/storage_attestation","attest_author":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM/action/author_attestation","sign_citation":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM/action/citation_signature","submit_replication":"https://pith.science/pith/MNNWJO3YLRMPWVA5EX2ADTZMPM/action/replication_record"}},"created_at":"2026-07-05T02:09:39.147418+00:00","updated_at":"2026-07-05T02:09:39.147418+00:00"}