{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:6HZPQNGDOJTIEJ5M5OEIWSOZ7R","short_pith_number":"pith:6HZPQNGD","canonical_record":{"source":{"id":"2605.14641","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:57:21Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6ff34a2526a747f9d1ade6be5c818ed2eb4f7cbc4f33e21b9b27d186ac15a3fe","abstract_canon_sha256":"fa4fe9d92c10cc2ee5c34b63bf138133d5268575f98f24d1709155486f664264"},"schema_version":"1.0"},"canonical_sha256":"f1f2f834c372668227aceb888b49d9fc6629c0b4701c77474f79655e86f449fd","source":{"kind":"arxiv","id":"2605.14641","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14641","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14641v1","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14641","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"pith_short_12","alias_value":"6HZPQNGDOJTI","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6HZPQNGDOJTIEJ5M","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6HZPQNGD","created_at":"2026-05-18T12:33:37Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:6HZPQNGDOJTIEJ5M5OEIWSOZ7R","target":"record","payload":{"canonical_record":{"source":{"id":"2605.14641","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:57:21Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"6ff34a2526a747f9d1ade6be5c818ed2eb4f7cbc4f33e21b9b27d186ac15a3fe","abstract_canon_sha256":"fa4fe9d92c10cc2ee5c34b63bf138133d5268575f98f24d1709155486f664264"},"schema_version":"1.0"},"canonical_sha256":"f1f2f834c372668227aceb888b49d9fc6629c0b4701c77474f79655e86f449fd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:39:03.886177Z","signature_b64":"bx1BLO1L4+opBycgr1QriJ0zr1WWUsRpd4PbNyKuQepTiS86/CwEUopBCtHfkSVuYqVWgsF0AlD6cfzdllkoBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f1f2f834c372668227aceb888b49d9fc6629c0b4701c77474f79655e86f449fd","last_reissued_at":"2026-05-17T23:39:03.885487Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:39:03.885487Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.14641","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"lM3neSdw5gZyjJMZRoKxMPsFE2XpFsF+GtXQ6ej4bvL8lmCJY9hY16vHrCXG5X/OfcKhb1681CrydVTV8HHXBw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T16:14:22.697365Z"},"content_sha256":"53048aa3ffaf2c91247694ca89d5e3ab04af9614140acf6ab474ce3d9a5ca33b","schema_version":"1.0","event_id":"sha256:53048aa3ffaf2c91247694ca89d5e3ab04af9614140acf6ab474ce3d9a5ca33b"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:6HZPQNGDOJTIEJ5M5OEIWSOZ7R","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How to Evaluate and Refine your CAM","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers.","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Alessandra Stramiglio, Luca Domeniconi, Michele Lombardi, Samuele Salti","submitted_at":"2026-05-14T09:57:21Z","abstract_excerpt":"Class attribution maps (CAMs) provide local explanations for the decisions of convolutional neural networks. While widely used in practice, the evaluation of CAMs remains challenging due to the lack of ground-truth explanations, making it difficult to evaluate the soundness of existing metrics. Independently, most commonly used CAM methods produce low-resolution attribution maps, which limits their usefulness for detailed interpretability.\n  To address the evaluation challenge, we introduce a synthetic dataset with ground-truth attributions that enables a rigorous comparison of CAM evaluation "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"RefineCAM consistently outperforms existing methods according to the proposed evaluation.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The synthetic dataset's ground-truth attributions accurately capture what constitutes faithful explanations in real-world image classification scenarios.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Introduces synthetic ground-truth dataset for CAM evaluation, proposes ARCC composite metric, and RefineCAM method that aggregates layers for higher-resolution maps outperforming baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"663a916edf062eae424e1913a32268c633a004cda6f60896644c7ca89152706d"},"source":{"id":"2605.14641","kind":"arxiv","version":1},"verdict":{"id":"8789065d-6482-46db-ae99-ee9e45697a80","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T05:38:43.919454Z","strongest_claim":"RefineCAM consistently outperforms existing methods according to the proposed evaluation.","one_line_summary":"Introduces synthetic ground-truth dataset for CAM evaluation, proposes ARCC composite metric, and RefineCAM method that aggregates layers for higher-resolution maps outperforming baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The synthetic dataset's ground-truth attributions accurately capture what constitutes faithful explanations in real-world image classification scenarios.","pith_extraction_headline":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers."},"references":{"count":32,"sample":[{"doi":"","year":2023,"title":"Nature Machine Intelligence5(9), 1006– 1019 (2023)","work_id":"7f5f0799-4350-45a3-9929-66ac014ad82f","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2021,"title":"Bohle, M., Fritz, M., Schiele, B.: Convolutional dynamic alignment networks for interpretableclassifications.In:ProceedingsoftheIEEE/CVFConferenceonCom- puter Vision and Pattern Recognition. pp. 10029","work_id":"d301994e-1914-426b-a1aa-c8519f9c4f9c","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1007/s00371-025-03803-1","year":2025,"title":"The Visual Computer 41(10), 7249–7267 (Jan 2025)","work_id":"f24ac45a-a861-419e-9095-cb0dd9433864","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"https://universe.roboflow.com/carddataset/ whereswaldy (2023)","work_id":"9baed613-7d02-4ab2-8bd3-a4d6db23fbb7","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2018,"title":"In: 2018 IEEE winter conference on applications of computer vision (WACV)","work_id":"0225a284-acca-44fe-9981-76614cc172c2","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":32,"snapshot_sha256":"2e4d1a2b92dfef8da2d294e90d082f6ef797c8f0e505c651badeb4ed2804be58","internal_anchors":3},"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"},"verdict_id":"8789065d-6482-46db-ae99-ee9e45697a80"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-17T23:39:03Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"XL73tu0GXqaz4EF8ob3cd0iMgbrb/glk9M/8JXqT3s0L41APDNAgi+sZgSmRMB78BwQQ+rhdVmdbei0LBmLFAA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-08-03T16:14:22.698227Z"},"content_sha256":"fa4362c3978cf8a8f9940f3532416473907d279d84b4e47e17232065c29aa9a5","schema_version":"1.0","event_id":"sha256:fa4362c3978cf8a8f9940f3532416473907d279d84b4e47e17232065c29aa9a5"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/bundle.json","state_url":"https://pith.science/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-08-03T16:14:22Z","links":{"resolver":"https://pith.science/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R","bundle":"https://pith.science/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/bundle.json","state":"https://pith.science/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/state.json","well_known_bundle":"https://pith.science/.well-known/pith/6HZPQNGDOJTIEJ5M5OEIWSOZ7R/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:6HZPQNGDOJTIEJ5M5OEIWSOZ7R","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"fa4fe9d92c10cc2ee5c34b63bf138133d5268575f98f24d1709155486f664264","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:57:21Z","title_canon_sha256":"6ff34a2526a747f9d1ade6be5c818ed2eb4f7cbc4f33e21b9b27d186ac15a3fe"},"schema_version":"1.0","source":{"id":"2605.14641","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.14641","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"arxiv_version","alias_value":"2605.14641v1","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14641","created_at":"2026-05-17T23:39:03Z"},{"alias_kind":"pith_short_12","alias_value":"6HZPQNGDOJTI","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_16","alias_value":"6HZPQNGDOJTIEJ5M","created_at":"2026-05-18T12:33:37Z"},{"alias_kind":"pith_short_8","alias_value":"6HZPQNGD","created_at":"2026-05-18T12:33:37Z"}],"graph_snapshots":[{"event_id":"sha256:fa4362c3978cf8a8f9940f3532416473907d279d84b4e47e17232065c29aa9a5","target":"graph","created_at":"2026-05-17T23:39:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"RefineCAM consistently outperforms existing methods according to the proposed evaluation."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"The synthetic dataset's ground-truth attributions accurately capture what constitutes faithful explanations in real-world image classification scenarios."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Introduces synthetic ground-truth dataset for CAM evaluation, proposes ARCC composite metric, and RefineCAM method that aggregates layers for higher-resolution maps outperforming baselines."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers."}],"snapshot_sha256":"663a916edf062eae424e1913a32268c633a004cda6f60896644c7ca89152706d"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"paper":{"abstract_excerpt":"Class attribution maps (CAMs) provide local explanations for the decisions of convolutional neural networks. While widely used in practice, the evaluation of CAMs remains challenging due to the lack of ground-truth explanations, making it difficult to evaluate the soundness of existing metrics. Independently, most commonly used CAM methods produce low-resolution attribution maps, which limits their usefulness for detailed interpretability.\n  To address the evaluation challenge, we introduce a synthetic dataset with ground-truth attributions that enables a rigorous comparison of CAM evaluation ","authors_text":"Alessandra Stramiglio, Luca Domeniconi, Michele Lombardi, Samuele Salti","cross_cats":["cs.AI"],"headline":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:57:21Z","title":"How to Evaluate and Refine your CAM"},"references":{"count":32,"internal_anchors":3,"resolved_work":32,"sample":[{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":1,"title":"Nature Machine Intelligence5(9), 1006– 1019 (2023)","work_id":"7f5f0799-4350-45a3-9929-66ac014ad82f","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":2,"title":"Bohle, M., Fritz, M., Schiele, B.: Convolutional dynamic alignment networks for interpretableclassifications.In:ProceedingsoftheIEEE/CVFConferenceonCom- puter Vision and Pattern Recognition. pp. 10029","work_id":"d301994e-1914-426b-a1aa-c8519f9c4f9c","year":2021},{"cited_arxiv_id":"","doi":"10.1007/s00371-025-03803-1","is_internal_anchor":false,"ref_index":3,"title":"The Visual Computer 41(10), 7249–7267 (Jan 2025)","work_id":"f24ac45a-a861-419e-9095-cb0dd9433864","year":2025},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":4,"title":"https://universe.roboflow.com/carddataset/ whereswaldy (2023)","work_id":"9baed613-7d02-4ab2-8bd3-a4d6db23fbb7","year":2023},{"cited_arxiv_id":"","doi":"","is_internal_anchor":false,"ref_index":5,"title":"In: 2018 IEEE winter conference on applications of computer vision (WACV)","work_id":"0225a284-acca-44fe-9981-76614cc172c2","year":2018}],"snapshot_sha256":"2e4d1a2b92dfef8da2d294e90d082f6ef797c8f0e505c651badeb4ed2804be58"},"source":{"id":"2605.14641","kind":"arxiv","version":1},"verdict":{"created_at":"2026-05-15T05:38:43.919454Z","id":"8789065d-6482-46db-ae99-ee9e45697a80","model_set":{"reader":"grok-4.3"},"one_line_summary":"Introduces synthetic ground-truth dataset for CAM evaluation, proposes ARCC composite metric, and RefineCAM method that aggregates layers for higher-resolution maps outperforming baselines.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"RefineCAM produces higher-resolution attribution maps for CNN decisions by aggregating multiple layers.","strongest_claim":"RefineCAM consistently outperforms existing methods according to the proposed evaluation.","weakest_assumption":"The synthetic dataset's ground-truth attributions accurately capture what constitutes faithful explanations in real-world image classification scenarios."}},"verdict_id":"8789065d-6482-46db-ae99-ee9e45697a80"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:53048aa3ffaf2c91247694ca89d5e3ab04af9614140acf6ab474ce3d9a5ca33b","target":"record","created_at":"2026-05-17T23:39:03Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"fa4fe9d92c10cc2ee5c34b63bf138133d5268575f98f24d1709155486f664264","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-05-14T09:57:21Z","title_canon_sha256":"6ff34a2526a747f9d1ade6be5c818ed2eb4f7cbc4f33e21b9b27d186ac15a3fe"},"schema_version":"1.0","source":{"id":"2605.14641","kind":"arxiv","version":1}},"canonical_sha256":"f1f2f834c372668227aceb888b49d9fc6629c0b4701c77474f79655e86f449fd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"f1f2f834c372668227aceb888b49d9fc6629c0b4701c77474f79655e86f449fd","first_computed_at":"2026-05-17T23:39:03.885487Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-17T23:39:03.885487Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"bx1BLO1L4+opBycgr1QriJ0zr1WWUsRpd4PbNyKuQepTiS86/CwEUopBCtHfkSVuYqVWgsF0AlD6cfzdllkoBA==","signature_status":"signed_v1","signed_at":"2026-05-17T23:39:03.886177Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.14641","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:53048aa3ffaf2c91247694ca89d5e3ab04af9614140acf6ab474ce3d9a5ca33b","sha256:fa4362c3978cf8a8f9940f3532416473907d279d84b4e47e17232065c29aa9a5"],"state_sha256":"3015a6a417870174a9a659b0a8a811995cc1bedcf51700b51d145e21153fc57b"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"i4p7UXI+l41C1Kk4qayDzx5We8HwZlrYVGWkoIrUX73SI94JUYpZ1kgw9amPkeO2+pSQBxjtfD+F8TylwjvMAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-08-03T16:14:22.705628Z","bundle_sha256":"18778d3a874ca4b7a13ead347e967f4beade798789085bbc173a97857e03145d"}}