{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:Z6T47CSGDOMF4ZLX4CDAPDFDDE","short_pith_number":"pith:Z6T47CSG","schema_version":"1.0","canonical_sha256":"cfa7cf8a461b985e6577e086078ca31936de7b312b77bf3474ca4bd65eb5f417","source":{"kind":"arxiv","id":"2605.22168","version":1},"attestation_state":"computed","paper":{"title":"Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Jo\\\"el Roman Ky, Maxime Cordy, Salah Ghamizi","submitted_at":"2026-05-21T08:39:46Z","abstract_excerpt":"Vision-Language Models (VLMs) map complex visual inputs to semantic spaces, but interpreting the cross-modal reasoning of VLMs currently relies on post-hoc explainers evaluated via unimodal perturbation metrics. We expose a limitation in this paradigm: because multimodal datasets contain language priors and modality biases, VLMs frequently exhibit cross-modal redundancy, allowing them to answer visual queries using text alone. Consequently, unimodal metrics penalize faithful explainers, triggering an evaluation collapse where visual and textual rankings fundamentally contradict each other. %(K"},"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":"2605.22168","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2026-05-21T08:39:46Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f5c99304da052f0fc765f150f84a6208db9d43245679cd8861597bf4de7f83b7","abstract_canon_sha256":"22e42d45624d27a949a774756299655224e9ab71c63e7fa83e25fac740267b03"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:04:29.808066Z","signature_b64":"vUtuf50DujquQgmlWdxmqFdvQOio6OqKFiQHICiCpBTn7/EiKIedgtqYlYJzHNLIkSepH9zhaSN1BFM/iSz8AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"cfa7cf8a461b985e6577e086078ca31936de7b312b77bf3474ca4bd65eb5f417","last_reissued_at":"2026-05-22T01:04:29.807267Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:04:29.807267Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Measuring Cross-Modal Synergy: A Benchmark for VLM Explainability","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Jo\\\"el Roman Ky, Maxime Cordy, Salah Ghamizi","submitted_at":"2026-05-21T08:39:46Z","abstract_excerpt":"Vision-Language Models (VLMs) map complex visual inputs to semantic spaces, but interpreting the cross-modal reasoning of VLMs currently relies on post-hoc explainers evaluated via unimodal perturbation metrics. We expose a limitation in this paradigm: because multimodal datasets contain language priors and modality biases, VLMs frequently exhibit cross-modal redundancy, allowing them to answer visual queries using text alone. Consequently, unimodal metrics penalize faithful explainers, triggering an evaluation collapse where visual and textual rankings fundamentally contradict each other. %(K"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.22168","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/2605.22168/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":"2605.22168","created_at":"2026-05-22T01:04:29.807423+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.22168v1","created_at":"2026-05-22T01:04:29.807423+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.22168","created_at":"2026-05-22T01:04:29.807423+00:00"},{"alias_kind":"pith_short_12","alias_value":"Z6T47CSGDOMF","created_at":"2026-05-22T01:04:29.807423+00:00"},{"alias_kind":"pith_short_16","alias_value":"Z6T47CSGDOMF4ZLX","created_at":"2026-05-22T01:04:29.807423+00:00"},{"alias_kind":"pith_short_8","alias_value":"Z6T47CSG","created_at":"2026-05-22T01:04:29.807423+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE","json":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE.json","graph_json":"https://pith.science/api/pith-number/Z6T47CSGDOMF4ZLX4CDAPDFDDE/graph.json","events_json":"https://pith.science/api/pith-number/Z6T47CSGDOMF4ZLX4CDAPDFDDE/events.json","paper":"https://pith.science/paper/Z6T47CSG"},"agent_actions":{"view_html":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE","download_json":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE.json","view_paper":"https://pith.science/paper/Z6T47CSG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.22168&json=true","fetch_graph":"https://pith.science/api/pith-number/Z6T47CSGDOMF4ZLX4CDAPDFDDE/graph.json","fetch_events":"https://pith.science/api/pith-number/Z6T47CSGDOMF4ZLX4CDAPDFDDE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE/action/storage_attestation","attest_author":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE/action/author_attestation","sign_citation":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE/action/citation_signature","submit_replication":"https://pith.science/pith/Z6T47CSGDOMF4ZLX4CDAPDFDDE/action/replication_record"}},"created_at":"2026-05-22T01:04:29.807423+00:00","updated_at":"2026-05-22T01:04:29.807423+00:00"}