{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:SBEKSNTKLQB2CB2LXKM25F4KKG","short_pith_number":"pith:SBEKSNTK","schema_version":"1.0","canonical_sha256":"9048a9366a5c03a1074bba99ae978a51bb6e0f944cb4af7ac1a241a8c8810593","source":{"kind":"arxiv","id":"2607.11436","version":1},"attestation_state":"computed","paper":{"title":"The Ebb and Flow of Multimodal Focus: Scheduling Visual Relay Windows for Grounded VLM Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Heng Tao Shen, Hongye Fang, Jingkuan Song, Sirui Da, Wencheng Ye, Xing Xu, Yi Bin, Yujuan Ding, Yun Zhang, Zheng Wang","submitted_at":"2026-07-13T11:44:23Z","abstract_excerpt":"Vision-language models increasingly succeed on multimodal reasoning benchmarks, yet their visual evidence often becomes unstable once it enters the language stack, weakening evidence-grounded reasoning. To understand this fragility, we examine the internal dynamics of VLMs through a mechanistic lens and uncover a stable three-stage redistribution of multimodal attention focus across depth: an early question-conditioned organization, a critical middle visual-dominant relay, and a late return to answer formation. We operationalize the middle phase as the Visual Relay Window (VRW), and show that "},"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":"2607.11436","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-07-13T11:44:23Z","cross_cats_sorted":[],"title_canon_sha256":"35bc24a8b62e94b815ee4d48461015afcca4fe1650ffb95a03939fce441dd8da","abstract_canon_sha256":"1869ef15a5eab11cf237c597eba4c00118b7cf67dfb8094aead8dcbb778ada92"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T02:22:02.704784Z","signature_b64":"LgYyikmDkmhz2S3eOmZBBIhHmURmXCyng0RFB+GtUy3YOqhCTVk36tvvcxNm+z1LD7B4KRTzAr98t60fOo6eBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9048a9366a5c03a1074bba99ae978a51bb6e0f944cb4af7ac1a241a8c8810593","last_reissued_at":"2026-07-14T02:22:02.703881Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T02:22:02.703881Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Ebb and Flow of Multimodal Focus: Scheduling Visual Relay Windows for Grounded VLM Reasoning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Heng Tao Shen, Hongye Fang, Jingkuan Song, Sirui Da, Wencheng Ye, Xing Xu, Yi Bin, Yujuan Ding, Yun Zhang, Zheng Wang","submitted_at":"2026-07-13T11:44:23Z","abstract_excerpt":"Vision-language models increasingly succeed on multimodal reasoning benchmarks, yet their visual evidence often becomes unstable once it enters the language stack, weakening evidence-grounded reasoning. To understand this fragility, we examine the internal dynamics of VLMs through a mechanistic lens and uncover a stable three-stage redistribution of multimodal attention focus across depth: an early question-conditioned organization, a critical middle visual-dominant relay, and a late return to answer formation. We operationalize the middle phase as the Visual Relay Window (VRW), and show that "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.11436","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/2607.11436/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":"2607.11436","created_at":"2026-07-14T02:22:02.704347+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.11436v1","created_at":"2026-07-14T02:22:02.704347+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.11436","created_at":"2026-07-14T02:22:02.704347+00:00"},{"alias_kind":"pith_short_12","alias_value":"SBEKSNTKLQB2","created_at":"2026-07-14T02:22:02.704347+00:00"},{"alias_kind":"pith_short_16","alias_value":"SBEKSNTKLQB2CB2L","created_at":"2026-07-14T02:22:02.704347+00:00"},{"alias_kind":"pith_short_8","alias_value":"SBEKSNTK","created_at":"2026-07-14T02:22:02.704347+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/SBEKSNTKLQB2CB2LXKM25F4KKG","json":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG.json","graph_json":"https://pith.science/api/pith-number/SBEKSNTKLQB2CB2LXKM25F4KKG/graph.json","events_json":"https://pith.science/api/pith-number/SBEKSNTKLQB2CB2LXKM25F4KKG/events.json","paper":"https://pith.science/paper/SBEKSNTK"},"agent_actions":{"view_html":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG","download_json":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG.json","view_paper":"https://pith.science/paper/SBEKSNTK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.11436&json=true","fetch_graph":"https://pith.science/api/pith-number/SBEKSNTKLQB2CB2LXKM25F4KKG/graph.json","fetch_events":"https://pith.science/api/pith-number/SBEKSNTKLQB2CB2LXKM25F4KKG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG/action/storage_attestation","attest_author":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG/action/author_attestation","sign_citation":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG/action/citation_signature","submit_replication":"https://pith.science/pith/SBEKSNTKLQB2CB2LXKM25F4KKG/action/replication_record"}},"created_at":"2026-07-14T02:22:02.704347+00:00","updated_at":"2026-07-14T02:22:02.704347+00:00"}