{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JXD6E7DDAMXZIMXKR2VSAVUUWT","short_pith_number":"pith:JXD6E7DD","schema_version":"1.0","canonical_sha256":"4dc7e27c63032f9432ea8eab205694b4d0744269ad189b80273818256cb5694c","source":{"kind":"arxiv","id":"2205.14865","version":4},"attestation_state":"computed","paper":{"title":"Prompt-aligned Gradient for Prompt Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Beier Zhu, Hanwang Zhang, Yucheng Han, Yue Wu, Yulei Niu","submitted_at":"2022-05-30T06:05:21Z","abstract_excerpt":"Thanks to the large pre-trained vision-language models (VLMs) like CLIP, we can craft a zero-shot classifier by \"prompt\", e.g., the confidence score of an image being \"[CLASS]\" can be obtained by using the VLM provided similarity measure between the image and the prompt sentence \"a photo of a [CLASS]\". Therefore, prompt shows a great potential for fast adaptation of VLMs to downstream tasks if we fine-tune the prompt-based similarity measure. However, we find a common failure that improper fine-tuning may not only undermine the prompt's inherent prediction for the task-related classes, but als"},"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":"2205.14865","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2022-05-30T06:05:21Z","cross_cats_sorted":[],"title_canon_sha256":"d90a8932fa12b3062c2ba1516efe225b49a217d22714111830282875657abcdc","abstract_canon_sha256":"6529c2e135f7881a2fa9b79e25dcd54635fb94fa53e09f77271e06646ae72131"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:52:50.692624Z","signature_b64":"A6CQPDKugsQQjDmrnOAQ6V0BBubkcwTpEDmGPOAxZ2B7kDXWc9Mnk+IBvIuxhCVHzNl0VBvSMm3SPLnu3jbqDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4dc7e27c63032f9432ea8eab205694b4d0744269ad189b80273818256cb5694c","last_reissued_at":"2026-07-05T11:52:50.692123Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:52:50.692123Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Prompt-aligned Gradient for Prompt Tuning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Beier Zhu, Hanwang Zhang, Yucheng Han, Yue Wu, Yulei Niu","submitted_at":"2022-05-30T06:05:21Z","abstract_excerpt":"Thanks to the large pre-trained vision-language models (VLMs) like CLIP, we can craft a zero-shot classifier by \"prompt\", e.g., the confidence score of an image being \"[CLASS]\" can be obtained by using the VLM provided similarity measure between the image and the prompt sentence \"a photo of a [CLASS]\". Therefore, prompt shows a great potential for fast adaptation of VLMs to downstream tasks if we fine-tune the prompt-based similarity measure. However, we find a common failure that improper fine-tuning may not only undermine the prompt's inherent prediction for the task-related classes, but als"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2205.14865","kind":"arxiv","version":4},"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/2205.14865/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":"2205.14865","created_at":"2026-07-05T11:52:50.692189+00:00"},{"alias_kind":"arxiv_version","alias_value":"2205.14865v4","created_at":"2026-07-05T11:52:50.692189+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2205.14865","created_at":"2026-07-05T11:52:50.692189+00:00"},{"alias_kind":"pith_short_12","alias_value":"JXD6E7DDAMXZ","created_at":"2026-07-05T11:52:50.692189+00:00"},{"alias_kind":"pith_short_16","alias_value":"JXD6E7DDAMXZIMXK","created_at":"2026-07-05T11:52:50.692189+00:00"},{"alias_kind":"pith_short_8","alias_value":"JXD6E7DD","created_at":"2026-07-05T11:52:50.692189+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.10575","citing_title":"Text to Image for Multi-Label Image Recognition with Joint Prompt-Adapter Learning","ref_index":58,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT","json":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT.json","graph_json":"https://pith.science/api/pith-number/JXD6E7DDAMXZIMXKR2VSAVUUWT/graph.json","events_json":"https://pith.science/api/pith-number/JXD6E7DDAMXZIMXKR2VSAVUUWT/events.json","paper":"https://pith.science/paper/JXD6E7DD"},"agent_actions":{"view_html":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT","download_json":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT.json","view_paper":"https://pith.science/paper/JXD6E7DD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2205.14865&json=true","fetch_graph":"https://pith.science/api/pith-number/JXD6E7DDAMXZIMXKR2VSAVUUWT/graph.json","fetch_events":"https://pith.science/api/pith-number/JXD6E7DDAMXZIMXKR2VSAVUUWT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT/action/storage_attestation","attest_author":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT/action/author_attestation","sign_citation":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT/action/citation_signature","submit_replication":"https://pith.science/pith/JXD6E7DDAMXZIMXKR2VSAVUUWT/action/replication_record"}},"created_at":"2026-07-05T11:52:50.692189+00:00","updated_at":"2026-07-05T11:52:50.692189+00:00"}