{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:25C7H22O4WSAKGQLER3VK7VROY","short_pith_number":"pith:25C7H22O","schema_version":"1.0","canonical_sha256":"d745f3eb4ee5a4051a0b2477557eb176294b7090b1ff147e1eb68b2e09f2f3b3","source":{"kind":"arxiv","id":"2606.12898","version":1},"attestation_state":"computed","paper":{"title":"Magnifying What Matters: Attention-Guided Adaptive Rendering for Visual Text Comprehension","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Hui Liu, Kai Guo, Qirui Wang, Shenglai Zeng, Xianxuan Long, Xinnan Dai","submitted_at":"2026-06-11T04:57:51Z","abstract_excerpt":"Visual Text Comprehension (VTC) renders text into images for a vision-language model (VLM) to read, sidestepping LLM context-window limits and powering applications from long-page OCR to multi-page memory QA. Yet existing VTC pipelines treat rendering and layout as a fixed, content-agnostic preprocessing step and offer little mechanistic understanding of how VLMs internally process visualized text. Through a focused empirical study on VTC QA tasks, we reveal that VLMs exhibit a localization-without-utilization regime: evidence-localizing attention emerges sharply in the middle-to-late layers a"},"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":"2606.12898","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-06-11T04:57:51Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"a3492896508a81b41f0b28a5042f7c3326100dbf2873b763827ad7262f7de74d","abstract_canon_sha256":"c4d54d44c76522cb6361bc3423b7f61e4829c1ec41d8bf200ddfce7c21cac821"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-12T01:09:31.799482Z","signature_b64":"Q8TxzVMtlKpl3uhYHPBA8xCtx7puqBQ7JIhiF4TsEa3zDCpb4PMD3Ha5uU2QzKME6L/Udjm1LSweXTeD8UpnCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d745f3eb4ee5a4051a0b2477557eb176294b7090b1ff147e1eb68b2e09f2f3b3","last_reissued_at":"2026-06-12T01:09:31.798994Z","signature_status":"signed_v1","first_computed_at":"2026-06-12T01:09:31.798994Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Magnifying What Matters: Attention-Guided Adaptive Rendering for Visual Text Comprehension","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Hui Liu, Kai Guo, Qirui Wang, Shenglai Zeng, Xianxuan Long, Xinnan Dai","submitted_at":"2026-06-11T04:57:51Z","abstract_excerpt":"Visual Text Comprehension (VTC) renders text into images for a vision-language model (VLM) to read, sidestepping LLM context-window limits and powering applications from long-page OCR to multi-page memory QA. Yet existing VTC pipelines treat rendering and layout as a fixed, content-agnostic preprocessing step and offer little mechanistic understanding of how VLMs internally process visualized text. Through a focused empirical study on VTC QA tasks, we reveal that VLMs exhibit a localization-without-utilization regime: evidence-localizing attention emerges sharply in the middle-to-late layers a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.12898","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/2606.12898/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":"2606.12898","created_at":"2026-06-12T01:09:31.799048+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.12898v1","created_at":"2026-06-12T01:09:31.799048+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.12898","created_at":"2026-06-12T01:09:31.799048+00:00"},{"alias_kind":"pith_short_12","alias_value":"25C7H22O4WSA","created_at":"2026-06-12T01:09:31.799048+00:00"},{"alias_kind":"pith_short_16","alias_value":"25C7H22O4WSAKGQL","created_at":"2026-06-12T01:09:31.799048+00:00"},{"alias_kind":"pith_short_8","alias_value":"25C7H22O","created_at":"2026-06-12T01:09:31.799048+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/25C7H22O4WSAKGQLER3VK7VROY","json":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY.json","graph_json":"https://pith.science/api/pith-number/25C7H22O4WSAKGQLER3VK7VROY/graph.json","events_json":"https://pith.science/api/pith-number/25C7H22O4WSAKGQLER3VK7VROY/events.json","paper":"https://pith.science/paper/25C7H22O"},"agent_actions":{"view_html":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY","download_json":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY.json","view_paper":"https://pith.science/paper/25C7H22O","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.12898&json=true","fetch_graph":"https://pith.science/api/pith-number/25C7H22O4WSAKGQLER3VK7VROY/graph.json","fetch_events":"https://pith.science/api/pith-number/25C7H22O4WSAKGQLER3VK7VROY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY/action/storage_attestation","attest_author":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY/action/author_attestation","sign_citation":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY/action/citation_signature","submit_replication":"https://pith.science/pith/25C7H22O4WSAKGQLER3VK7VROY/action/replication_record"}},"created_at":"2026-06-12T01:09:31.799048+00:00","updated_at":"2026-06-12T01:09:31.799048+00:00"}