{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:NI7VAKCBOFPKUBL5KEBE4OSRCI","short_pith_number":"pith:NI7VAKCB","schema_version":"1.0","canonical_sha256":"6a3f502841715eaa057d51024e3a511204d7c0c41eb8d4edda89527ff7ea9d37","source":{"kind":"arxiv","id":"2212.10562","version":2},"attestation_state":"computed","paper":{"title":"Character-Aware Models Improve Visual Text Rendering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Adam Roberts, Chitwan Saharia, Dan Garrette, Irina Blok, Mohammad Norouzi, Noah Constant, RJ Mical, Rosanne Liu, Sharan Narang, William Chan","submitted_at":"2022-12-20T18:59:23Z","abstract_excerpt":"Current image generation models struggle to reliably produce well-formed visual text. In this paper, we investigate a key contributing factor: popular text-to-image models lack character-level input features, making it much harder to predict a word's visual makeup as a series of glyphs. To quantify this effect, we conduct a series of experiments comparing character-aware vs. character-blind text encoders. In the text-only domain, we find that character-aware models provide large gains on a novel spelling task (WikiSpell). Applying our learnings to the visual domain, we train a suite of image g"},"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":"2212.10562","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2022-12-20T18:59:23Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"ce8e4aef6054a6ff8fb4c843fc5c260e5d3dae115f5c061dc624399069a7f85d","abstract_canon_sha256":"ff41af82bec2494a017eae49d0b11d9b564d025aaa987652b05923f638e863d2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:06:42.525984Z","signature_b64":"AT4LQ/IngXXRUzDosPvD+Sq2PPYeSFPayUFUgipElYVFSmtw7lb74NdcOju+j7/h4TRrLRuuuRM5iDsPK3CgDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6a3f502841715eaa057d51024e3a511204d7c0c41eb8d4edda89527ff7ea9d37","last_reissued_at":"2026-07-05T06:06:42.525489Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:06:42.525489Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Character-Aware Models Improve Visual Text Rendering","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Adam Roberts, Chitwan Saharia, Dan Garrette, Irina Blok, Mohammad Norouzi, Noah Constant, RJ Mical, Rosanne Liu, Sharan Narang, William Chan","submitted_at":"2022-12-20T18:59:23Z","abstract_excerpt":"Current image generation models struggle to reliably produce well-formed visual text. In this paper, we investigate a key contributing factor: popular text-to-image models lack character-level input features, making it much harder to predict a word's visual makeup as a series of glyphs. To quantify this effect, we conduct a series of experiments comparing character-aware vs. character-blind text encoders. In the text-only domain, we find that character-aware models provide large gains on a novel spelling task (WikiSpell). Applying our learnings to the visual domain, we train a suite of image g"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.10562","kind":"arxiv","version":2},"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/2212.10562/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":"2212.10562","created_at":"2026-07-05T06:06:42.525547+00:00"},{"alias_kind":"arxiv_version","alias_value":"2212.10562v2","created_at":"2026-07-05T06:06:42.525547+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.10562","created_at":"2026-07-05T06:06:42.525547+00:00"},{"alias_kind":"pith_short_12","alias_value":"NI7VAKCBOFPK","created_at":"2026-07-05T06:06:42.525547+00:00"},{"alias_kind":"pith_short_16","alias_value":"NI7VAKCBOFPKUBL5","created_at":"2026-07-05T06:06:42.525547+00:00"},{"alias_kind":"pith_short_8","alias_value":"NI7VAKCB","created_at":"2026-07-05T06:06:42.525547+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.24171","citing_title":"POCA: Pareto-Optimal Curriculum Alignment for Visual Text Generation","ref_index":22,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI","json":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI.json","graph_json":"https://pith.science/api/pith-number/NI7VAKCBOFPKUBL5KEBE4OSRCI/graph.json","events_json":"https://pith.science/api/pith-number/NI7VAKCBOFPKUBL5KEBE4OSRCI/events.json","paper":"https://pith.science/paper/NI7VAKCB"},"agent_actions":{"view_html":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI","download_json":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI.json","view_paper":"https://pith.science/paper/NI7VAKCB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2212.10562&json=true","fetch_graph":"https://pith.science/api/pith-number/NI7VAKCBOFPKUBL5KEBE4OSRCI/graph.json","fetch_events":"https://pith.science/api/pith-number/NI7VAKCBOFPKUBL5KEBE4OSRCI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI/action/storage_attestation","attest_author":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI/action/author_attestation","sign_citation":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI/action/citation_signature","submit_replication":"https://pith.science/pith/NI7VAKCBOFPKUBL5KEBE4OSRCI/action/replication_record"}},"created_at":"2026-07-05T06:06:42.525547+00:00","updated_at":"2026-07-05T06:06:42.525547+00:00"}