{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:7RPDDNGQOCPUPLIY4LBJCD7EFW","short_pith_number":"pith:7RPDDNGQ","schema_version":"1.0","canonical_sha256":"fc5e31b4d0709f47ad18e2c2910fe42d9980356cdee3faff9c5e0bcdc7ef6d7d","source":{"kind":"arxiv","id":"2205.03720","version":2},"attestation_state":"computed","paper":{"title":"Empowering parameter-efficient transfer learning by recognizing the kernel structure in self-attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Devamanyu Hazarika, Di Jin, Dilek Hakkani-Tur, Mahdi Namazifar, Yang Liu, Yifan Chen","submitted_at":"2022-05-07T20:52:54Z","abstract_excerpt":"The massive amount of trainable parameters in the pre-trained language models (PLMs) makes them hard to be deployed to multiple downstream tasks. To address this issue, parameter-efficient transfer learning methods have been proposed to tune only a few parameters during fine-tuning while freezing the rest. This paper looks at existing methods along this line through the \\textit{kernel lens}. Motivated by the connection between self-attention in transformer-based PLMs and kernel learning, we propose \\textit{kernel-wise adapters}, namely \\textit{Kernel-mix}, that utilize the kernel structure in "},"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":"2205.03720","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-05-07T20:52:54Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"526e882c9f068ea5548745dc3a6e4cc9333d1d49120cf91ffce87a6312cc85e9","abstract_canon_sha256":"2aefdd2fa25c4f26bb58d055257eb067c1981b08d2ba0e85c48ac9cc7eae9a6b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:10:37.698622Z","signature_b64":"/zRjZn4WhCVQO+cYp1E8JW9Am32PH7n4gpE+skXcPqqZtX4TTWWkY2+R8SXXY13ABRu1yMAiUyx4b/8mqm5jAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"fc5e31b4d0709f47ad18e2c2910fe42d9980356cdee3faff9c5e0bcdc7ef6d7d","last_reissued_at":"2026-07-05T05:10:37.698084Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:10:37.698084Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Empowering parameter-efficient transfer learning by recognizing the kernel structure in self-attention","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Devamanyu Hazarika, Di Jin, Dilek Hakkani-Tur, Mahdi Namazifar, Yang Liu, Yifan Chen","submitted_at":"2022-05-07T20:52:54Z","abstract_excerpt":"The massive amount of trainable parameters in the pre-trained language models (PLMs) makes them hard to be deployed to multiple downstream tasks. To address this issue, parameter-efficient transfer learning methods have been proposed to tune only a few parameters during fine-tuning while freezing the rest. This paper looks at existing methods along this line through the \\textit{kernel lens}. Motivated by the connection between self-attention in transformer-based PLMs and kernel learning, we propose \\textit{kernel-wise adapters}, namely \\textit{Kernel-mix}, that utilize the kernel structure in "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2205.03720","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/2205.03720/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":"2205.03720","created_at":"2026-07-05T05:10:37.698149+00:00"},{"alias_kind":"arxiv_version","alias_value":"2205.03720v2","created_at":"2026-07-05T05:10:37.698149+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2205.03720","created_at":"2026-07-05T05:10:37.698149+00:00"},{"alias_kind":"pith_short_12","alias_value":"7RPDDNGQOCPU","created_at":"2026-07-05T05:10:37.698149+00:00"},{"alias_kind":"pith_short_16","alias_value":"7RPDDNGQOCPUPLIY","created_at":"2026-07-05T05:10:37.698149+00:00"},{"alias_kind":"pith_short_8","alias_value":"7RPDDNGQ","created_at":"2026-07-05T05:10:37.698149+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW","json":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW.json","graph_json":"https://pith.science/api/pith-number/7RPDDNGQOCPUPLIY4LBJCD7EFW/graph.json","events_json":"https://pith.science/api/pith-number/7RPDDNGQOCPUPLIY4LBJCD7EFW/events.json","paper":"https://pith.science/paper/7RPDDNGQ"},"agent_actions":{"view_html":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW","download_json":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW.json","view_paper":"https://pith.science/paper/7RPDDNGQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2205.03720&json=true","fetch_graph":"https://pith.science/api/pith-number/7RPDDNGQOCPUPLIY4LBJCD7EFW/graph.json","fetch_events":"https://pith.science/api/pith-number/7RPDDNGQOCPUPLIY4LBJCD7EFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW/action/storage_attestation","attest_author":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW/action/author_attestation","sign_citation":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW/action/citation_signature","submit_replication":"https://pith.science/pith/7RPDDNGQOCPUPLIY4LBJCD7EFW/action/replication_record"}},"created_at":"2026-07-05T05:10:37.698149+00:00","updated_at":"2026-07-05T05:10:37.698149+00:00"}