{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:A3TJBGVGMRIV3VV2W73UIUB64K","short_pith_number":"pith:A3TJBGVG","schema_version":"1.0","canonical_sha256":"06e6909aa664515dd6bab7f744503ee2988d665a170d7b22aa04647751c971d6","source":{"kind":"arxiv","id":"2305.02423","version":1},"attestation_state":"computed","paper":{"title":"PTP: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Heng Huang, Lichang Chen, Minhao Cheng","submitted_at":"2023-05-03T20:30:51Z","abstract_excerpt":"Recent studies show that prompt tuning can better leverage the power of large language models than fine-tuning on downstream natural language understanding tasks. However, the existing prompt tuning methods have training instability issues, as the variance of scores under different random seeds is quite large. To address this critical problem, we first investigate and find that the loss landscape of vanilla prompt tuning is precipitous when it is visualized, where a slight change of input data can cause a big fluctuation in the loss landscape. This is an essential factor that leads to the inst"},"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":"2305.02423","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-03T20:30:51Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6e1c9a6c023b279fb40c6c1479bcfc8668810b76361c30a931b167e29bfc270a","abstract_canon_sha256":"47fc64b87edf58392cd958c9ce5d177d755d2a89d71c1ebff574b32cb545d81c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:07:00.658429Z","signature_b64":"FwlDyimsQPDhhrRCWxKnSLI4DhhY/eWu5IW1dpBVZ10gTe3XE4irY3JMQNclEzvOTKowcuSvCHsF9HU1G34vDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"06e6909aa664515dd6bab7f744503ee2988d665a170d7b22aa04647751c971d6","last_reissued_at":"2026-07-05T06:07:00.657764Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:07:00.657764Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PTP: Boosting Stability and Performance of Prompt Tuning with Perturbation-Based Regularizer","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Heng Huang, Lichang Chen, Minhao Cheng","submitted_at":"2023-05-03T20:30:51Z","abstract_excerpt":"Recent studies show that prompt tuning can better leverage the power of large language models than fine-tuning on downstream natural language understanding tasks. However, the existing prompt tuning methods have training instability issues, as the variance of scores under different random seeds is quite large. To address this critical problem, we first investigate and find that the loss landscape of vanilla prompt tuning is precipitous when it is visualized, where a slight change of input data can cause a big fluctuation in the loss landscape. This is an essential factor that leads to the inst"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.02423","kind":"arxiv","version":1},"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/2305.02423/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":"2305.02423","created_at":"2026-07-05T06:07:00.657825+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.02423v1","created_at":"2026-07-05T06:07:00.657825+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.02423","created_at":"2026-07-05T06:07:00.657825+00:00"},{"alias_kind":"pith_short_12","alias_value":"A3TJBGVGMRIV","created_at":"2026-07-05T06:07:00.657825+00:00"},{"alias_kind":"pith_short_16","alias_value":"A3TJBGVGMRIV3VV2","created_at":"2026-07-05T06:07:00.657825+00:00"},{"alias_kind":"pith_short_8","alias_value":"A3TJBGVG","created_at":"2026-07-05T06:07:00.657825+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.10488","citing_title":"5% > 100%: Flatness Preference is All You Need for Multimodal Parameter-Efficient Fine-Tuning","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2403.14608","citing_title":"Parameter-Efficient Fine-Tuning for Large Models: A Comprehensive Survey","ref_index":55,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K","json":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K.json","graph_json":"https://pith.science/api/pith-number/A3TJBGVGMRIV3VV2W73UIUB64K/graph.json","events_json":"https://pith.science/api/pith-number/A3TJBGVGMRIV3VV2W73UIUB64K/events.json","paper":"https://pith.science/paper/A3TJBGVG"},"agent_actions":{"view_html":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K","download_json":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K.json","view_paper":"https://pith.science/paper/A3TJBGVG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.02423&json=true","fetch_graph":"https://pith.science/api/pith-number/A3TJBGVGMRIV3VV2W73UIUB64K/graph.json","fetch_events":"https://pith.science/api/pith-number/A3TJBGVGMRIV3VV2W73UIUB64K/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K/action/storage_attestation","attest_author":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K/action/author_attestation","sign_citation":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K/action/citation_signature","submit_replication":"https://pith.science/pith/A3TJBGVGMRIV3VV2W73UIUB64K/action/replication_record"}},"created_at":"2026-07-05T06:07:00.657825+00:00","updated_at":"2026-07-05T06:07:00.657825+00:00"}