{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:B5LLNFVRTJED237ZDN7KL4WRB4","short_pith_number":"pith:B5LLNFVR","schema_version":"1.0","canonical_sha256":"0f56b696b19a483d6ff91b7ea5f2d10f02463b76403127ae3b9b3e74dd65c165","source":{"kind":"arxiv","id":"2504.17023","version":1},"attestation_state":"computed","paper":{"title":"What Makes for a Good Saliency Map? Comparing Strategies for Evaluating Saliency Maps in Explainable AI (XAI)","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.HC","authors_text":"Felix Kares, Hanwei Zhang, Markus Langer, Timo Speith","submitted_at":"2025-04-23T18:09:06Z","abstract_excerpt":"Saliency maps are a popular approach for explaining classifications of (convolutional) neural networks. However, it remains an open question as to how best to evaluate salience maps, with three families of evaluation methods commonly being used: subjective user measures, objective user measures, and mathematical metrics. We examine three of the most popular saliency map approaches (viz., LIME, Grad-CAM, and Guided Backpropagation) in a between subject study (N=166) across these families of evaluation methods. We test 1) for subjective measures, if the maps differ with respect to user trust and"},"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":"2504.17023","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.HC","submitted_at":"2025-04-23T18:09:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"41f18e97147afc85ed87a592b6fe9c39426250b709ad54642e3e575761ad4041","abstract_canon_sha256":"0a25915f253fa0115aa42d700ab917ff40deafabd8c849d2836e68f3955d7f83"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:53:20.714751Z","signature_b64":"m4kp1pcMlk+G3UtX/0+4mW+iLrjViE4Q7LMlkceU2bdwfRlblskhUNNHD4JxMMW6Oq3WVf6itsG/aSdHBTN8CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0f56b696b19a483d6ff91b7ea5f2d10f02463b76403127ae3b9b3e74dd65c165","last_reissued_at":"2026-07-05T10:53:20.714253Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:53:20.714253Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"What Makes for a Good Saliency Map? Comparing Strategies for Evaluating Saliency Maps in Explainable AI (XAI)","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.HC","authors_text":"Felix Kares, Hanwei Zhang, Markus Langer, Timo Speith","submitted_at":"2025-04-23T18:09:06Z","abstract_excerpt":"Saliency maps are a popular approach for explaining classifications of (convolutional) neural networks. However, it remains an open question as to how best to evaluate salience maps, with three families of evaluation methods commonly being used: subjective user measures, objective user measures, and mathematical metrics. We examine three of the most popular saliency map approaches (viz., LIME, Grad-CAM, and Guided Backpropagation) in a between subject study (N=166) across these families of evaluation methods. We test 1) for subjective measures, if the maps differ with respect to user trust and"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.17023","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/2504.17023/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":"2504.17023","created_at":"2026-07-05T10:53:20.714310+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.17023v1","created_at":"2026-07-05T10:53:20.714310+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.17023","created_at":"2026-07-05T10:53:20.714310+00:00"},{"alias_kind":"pith_short_12","alias_value":"B5LLNFVRTJED","created_at":"2026-07-05T10:53:20.714310+00:00"},{"alias_kind":"pith_short_16","alias_value":"B5LLNFVRTJED237Z","created_at":"2026-07-05T10:53:20.714310+00:00"},{"alias_kind":"pith_short_8","alias_value":"B5LLNFVR","created_at":"2026-07-05T10:53:20.714310+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07316","citing_title":"Mechanistic Interpretability for Neural Networks: Circuits, Sparse Features and Symbolic Reasoning","ref_index":4,"is_internal_anchor":true},{"citing_arxiv_id":"2605.20081","citing_title":"Bridging the Disciplinary Gap in Explainable AI: From Abstract Desiderata to Concrete Tasks","ref_index":37,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4","json":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4.json","graph_json":"https://pith.science/api/pith-number/B5LLNFVRTJED237ZDN7KL4WRB4/graph.json","events_json":"https://pith.science/api/pith-number/B5LLNFVRTJED237ZDN7KL4WRB4/events.json","paper":"https://pith.science/paper/B5LLNFVR"},"agent_actions":{"view_html":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4","download_json":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4.json","view_paper":"https://pith.science/paper/B5LLNFVR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.17023&json=true","fetch_graph":"https://pith.science/api/pith-number/B5LLNFVRTJED237ZDN7KL4WRB4/graph.json","fetch_events":"https://pith.science/api/pith-number/B5LLNFVRTJED237ZDN7KL4WRB4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4/action/storage_attestation","attest_author":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4/action/author_attestation","sign_citation":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4/action/citation_signature","submit_replication":"https://pith.science/pith/B5LLNFVRTJED237ZDN7KL4WRB4/action/replication_record"}},"created_at":"2026-07-05T10:53:20.714310+00:00","updated_at":"2026-07-05T10:53:20.714310+00:00"}