{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:NKMPEODNJSGBLVYUSHK2CRIVP3","short_pith_number":"pith:NKMPEODN","schema_version":"1.0","canonical_sha256":"6a98f2386d4c8c15d71491d5a145157ecdebb9ebdc10b8e02794d54f54bb4625","source":{"kind":"arxiv","id":"2605.19320","version":1},"attestation_state":"computed","paper":{"title":"TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CV","authors_text":"Fajri Koto, Fengxian Ji, Jiaming Wang, Jingpu Yang, Mingxuan Cui, Qian Jiang, Xiuying Chen, Zhecheng Shi, Zirui Song","submitted_at":"2026-05-19T03:55:59Z","abstract_excerpt":"Faithful text rendering remains a persistent weakness of large text-to-image generative models, as it requires both semantic instruction following and fine-grained glyph-level structure. Prior methods often improve this ability through architecture-specific modules or encoder modifications, which complicate deployment across foundation models. We study text rendering as a post-training preference-alignment problem and propose TextAlign, a non-invasive framework that keeps the generator architecture unchanged. The key component is a hierarchical vision-language model (VLM)-based reward that dec"},"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":"2605.19320","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-05-19T03:55:59Z","cross_cats_sorted":["cs.DB"],"title_canon_sha256":"845f6187752a0027505ac38e183f44d4a1b15701bd6a414ae290bd882845151e","abstract_canon_sha256":"1887168b9b640e721f06ce5b701a044571c8307b5e98334204b19e388f7d130f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T01:05:39.160931Z","signature_b64":"sXVXw6tEnLD/HoqU3DZGHWwQQlV2qFgSdohBum4CPVd26Ncq9Cnm1Aowvh2ZpCWrIghnqmr9XeGw7jh82rsrDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6a98f2386d4c8c15d71491d5a145157ecdebb9ebdc10b8e02794d54f54bb4625","last_reissued_at":"2026-05-20T01:05:39.160227Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T01:05:39.160227Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TextAlign: Preference Alignment for Text Rendering with Hierarchical Rewards","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.DB"],"primary_cat":"cs.CV","authors_text":"Fajri Koto, Fengxian Ji, Jiaming Wang, Jingpu Yang, Mingxuan Cui, Qian Jiang, Xiuying Chen, Zhecheng Shi, Zirui Song","submitted_at":"2026-05-19T03:55:59Z","abstract_excerpt":"Faithful text rendering remains a persistent weakness of large text-to-image generative models, as it requires both semantic instruction following and fine-grained glyph-level structure. Prior methods often improve this ability through architecture-specific modules or encoder modifications, which complicate deployment across foundation models. We study text rendering as a post-training preference-alignment problem and propose TextAlign, a non-invasive framework that keeps the generator architecture unchanged. The key component is a hierarchical vision-language model (VLM)-based reward that dec"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.19320","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/2605.19320/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":"2605.19320","created_at":"2026-05-20T01:05:39.160351+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.19320v1","created_at":"2026-05-20T01:05:39.160351+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.19320","created_at":"2026-05-20T01:05:39.160351+00:00"},{"alias_kind":"pith_short_12","alias_value":"NKMPEODNJSGB","created_at":"2026-05-20T01:05:39.160351+00:00"},{"alias_kind":"pith_short_16","alias_value":"NKMPEODNJSGBLVYU","created_at":"2026-05-20T01:05:39.160351+00:00"},{"alias_kind":"pith_short_8","alias_value":"NKMPEODN","created_at":"2026-05-20T01:05:39.160351+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/NKMPEODNJSGBLVYUSHK2CRIVP3","json":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3.json","graph_json":"https://pith.science/api/pith-number/NKMPEODNJSGBLVYUSHK2CRIVP3/graph.json","events_json":"https://pith.science/api/pith-number/NKMPEODNJSGBLVYUSHK2CRIVP3/events.json","paper":"https://pith.science/paper/NKMPEODN"},"agent_actions":{"view_html":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3","download_json":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3.json","view_paper":"https://pith.science/paper/NKMPEODN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.19320&json=true","fetch_graph":"https://pith.science/api/pith-number/NKMPEODNJSGBLVYUSHK2CRIVP3/graph.json","fetch_events":"https://pith.science/api/pith-number/NKMPEODNJSGBLVYUSHK2CRIVP3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3/action/storage_attestation","attest_author":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3/action/author_attestation","sign_citation":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3/action/citation_signature","submit_replication":"https://pith.science/pith/NKMPEODNJSGBLVYUSHK2CRIVP3/action/replication_record"}},"created_at":"2026-05-20T01:05:39.160351+00:00","updated_at":"2026-05-20T01:05:39.160351+00:00"}