{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:ZVPDL6LFIKTJXDMYOVTO5FQXNS","short_pith_number":"pith:ZVPDL6LF","schema_version":"1.0","canonical_sha256":"cd5e35f96542a69b8d987566ee96176c9c1ab28f949579fb866399983badcd9d","source":{"kind":"arxiv","id":"2206.01161","version":1},"attestation_state":"computed","paper":{"title":"Optimizing Relevance Maps of Vision Transformers Improves Robustness","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Hila Chefer, Idan Schwartz, Lior Wolf","submitted_at":"2022-06-02T17:24:48Z","abstract_excerpt":"It has been observed that visual classification models often rely mostly on the image background, neglecting the foreground, which hurts their robustness to distribution changes. To alleviate this shortcoming, we propose to monitor the model's relevancy signal and manipulate it such that the model is focused on the foreground object. This is done as a finetuning step, involving relatively few samples consisting of pairs of images and their associated foreground masks. Specifically, we encourage the model's relevancy map (i) to assign lower relevance to background regions, (ii) to consider as m"},"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":"2206.01161","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2022-06-02T17:24:48Z","cross_cats_sorted":[],"title_canon_sha256":"d682ffcdd616857290441edde9059e37c001957706c276febfc9265be9ec88ca","abstract_canon_sha256":"2322d89a28803e1915ee427985dffc2de6c601394f29596890cf1d35a65a9ee7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:28:37.571030Z","signature_b64":"b7pF4+ili0KXNua86esLzwNeJCYm7AKbiiAWkzGKBkxNF+2sxxP/G1RYjXcbBfgEJOOo1ZeW2lV0tTcb2AVZAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cd5e35f96542a69b8d987566ee96176c9c1ab28f949579fb866399983badcd9d","last_reissued_at":"2026-07-05T04:28:37.570406Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:28:37.570406Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Optimizing Relevance Maps of Vision Transformers Improves Robustness","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Hila Chefer, Idan Schwartz, Lior Wolf","submitted_at":"2022-06-02T17:24:48Z","abstract_excerpt":"It has been observed that visual classification models often rely mostly on the image background, neglecting the foreground, which hurts their robustness to distribution changes. To alleviate this shortcoming, we propose to monitor the model's relevancy signal and manipulate it such that the model is focused on the foreground object. This is done as a finetuning step, involving relatively few samples consisting of pairs of images and their associated foreground masks. Specifically, we encourage the model's relevancy map (i) to assign lower relevance to background regions, (ii) to consider as m"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2206.01161","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/2206.01161/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":"2206.01161","created_at":"2026-07-05T04:28:37.570499+00:00"},{"alias_kind":"arxiv_version","alias_value":"2206.01161v1","created_at":"2026-07-05T04:28:37.570499+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2206.01161","created_at":"2026-07-05T04:28:37.570499+00:00"},{"alias_kind":"pith_short_12","alias_value":"ZVPDL6LFIKTJ","created_at":"2026-07-05T04:28:37.570499+00:00"},{"alias_kind":"pith_short_16","alias_value":"ZVPDL6LFIKTJXDMY","created_at":"2026-07-05T04:28:37.570499+00:00"},{"alias_kind":"pith_short_8","alias_value":"ZVPDL6LF","created_at":"2026-07-05T04:28:37.570499+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2507.04788","citing_title":"Machine Learning from Explanations","ref_index":2,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS","json":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS.json","graph_json":"https://pith.science/api/pith-number/ZVPDL6LFIKTJXDMYOVTO5FQXNS/graph.json","events_json":"https://pith.science/api/pith-number/ZVPDL6LFIKTJXDMYOVTO5FQXNS/events.json","paper":"https://pith.science/paper/ZVPDL6LF"},"agent_actions":{"view_html":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS","download_json":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS.json","view_paper":"https://pith.science/paper/ZVPDL6LF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2206.01161&json=true","fetch_graph":"https://pith.science/api/pith-number/ZVPDL6LFIKTJXDMYOVTO5FQXNS/graph.json","fetch_events":"https://pith.science/api/pith-number/ZVPDL6LFIKTJXDMYOVTO5FQXNS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS/action/storage_attestation","attest_author":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS/action/author_attestation","sign_citation":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS/action/citation_signature","submit_replication":"https://pith.science/pith/ZVPDL6LFIKTJXDMYOVTO5FQXNS/action/replication_record"}},"created_at":"2026-07-05T04:28:37.570499+00:00","updated_at":"2026-07-05T04:28:37.570499+00:00"}