{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:K4IXYVUCXXMBSXGRDX6JH4LWG5","short_pith_number":"pith:K4IXYVUC","schema_version":"1.0","canonical_sha256":"57117c5682bdd8195cd11dfc93f1763768aed348b97292ac948e48d3561e7b0b","source":{"kind":"arxiv","id":"2410.05249","version":5},"attestation_state":"computed","paper":{"title":"LoTLIP: Improving Language-Image Pre-training for Long Text Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fan Lu, Kecheng Zheng, Qingpei Guo, Shuailei Ma, Wei Chen, Wei Wu, Yifei Zhang, Yujun Shen, Yuxin Guo, Zheng-Jun Zha","submitted_at":"2024-10-07T17:52:56Z","abstract_excerpt":"Understanding long text is of great demands in practice but beyond the reach of most language-image pre-training (LIP) models. In this work, we empirically confirm that the key reason causing such an issue is that the training images are usually paired with short captions, leaving certain tokens easily overshadowed by salient tokens. Towards this problem, our initial attempt is to relabel the data with long captions, however, directly learning with which may lead to performance degradation in understanding short text (e.g., in the image classification task). Then, after incorporating corner to"},"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":"2410.05249","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-10-07T17:52:56Z","cross_cats_sorted":[],"title_canon_sha256":"6df378060632287ea3b420443ff0e30e457d0ee3e751ca433403958f105e52c1","abstract_canon_sha256":"90edb7a9cd0175f90b7b060dc577f97be3aecf6a408b022eadebb9de22a56868"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:34:38.679428Z","signature_b64":"7+jtF/a7Gdmjx3Ob2GWOBhDJbPuVvRcINy1jvY1pBGMT5iI2tIwqCWRbrKudw/3PtpKuU+iW1WjAYY3KT+PJDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"57117c5682bdd8195cd11dfc93f1763768aed348b97292ac948e48d3561e7b0b","last_reissued_at":"2026-07-05T09:34:38.678928Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:34:38.678928Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoTLIP: Improving Language-Image Pre-training for Long Text Understanding","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Fan Lu, Kecheng Zheng, Qingpei Guo, Shuailei Ma, Wei Chen, Wei Wu, Yifei Zhang, Yujun Shen, Yuxin Guo, Zheng-Jun Zha","submitted_at":"2024-10-07T17:52:56Z","abstract_excerpt":"Understanding long text is of great demands in practice but beyond the reach of most language-image pre-training (LIP) models. In this work, we empirically confirm that the key reason causing such an issue is that the training images are usually paired with short captions, leaving certain tokens easily overshadowed by salient tokens. Towards this problem, our initial attempt is to relabel the data with long captions, however, directly learning with which may lead to performance degradation in understanding short text (e.g., in the image classification task). Then, after incorporating corner to"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.05249","kind":"arxiv","version":5},"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/2410.05249/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":"2410.05249","created_at":"2026-07-05T09:34:38.678989+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.05249v5","created_at":"2026-07-05T09:34:38.678989+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.05249","created_at":"2026-07-05T09:34:38.678989+00:00"},{"alias_kind":"pith_short_12","alias_value":"K4IXYVUCXXMB","created_at":"2026-07-05T09:34:38.678989+00:00"},{"alias_kind":"pith_short_16","alias_value":"K4IXYVUCXXMBSXGR","created_at":"2026-07-05T09:34:38.678989+00:00"},{"alias_kind":"pith_short_8","alias_value":"K4IXYVUC","created_at":"2026-07-05T09:34:38.678989+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2509.19207","citing_title":"Long Story Short: Disentangling Compositionality and Long-Caption Understanding in Contrastive VLMs","ref_index":42,"is_internal_anchor":false},{"citing_arxiv_id":"2509.23370","citing_title":"GRAPE: Let GRPO Supervise Query Rewriting by Ranking for Retrieval","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5","json":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5.json","graph_json":"https://pith.science/api/pith-number/K4IXYVUCXXMBSXGRDX6JH4LWG5/graph.json","events_json":"https://pith.science/api/pith-number/K4IXYVUCXXMBSXGRDX6JH4LWG5/events.json","paper":"https://pith.science/paper/K4IXYVUC"},"agent_actions":{"view_html":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5","download_json":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5.json","view_paper":"https://pith.science/paper/K4IXYVUC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.05249&json=true","fetch_graph":"https://pith.science/api/pith-number/K4IXYVUCXXMBSXGRDX6JH4LWG5/graph.json","fetch_events":"https://pith.science/api/pith-number/K4IXYVUCXXMBSXGRDX6JH4LWG5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5/action/storage_attestation","attest_author":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5/action/author_attestation","sign_citation":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5/action/citation_signature","submit_replication":"https://pith.science/pith/K4IXYVUCXXMBSXGRDX6JH4LWG5/action/replication_record"}},"created_at":"2026-07-05T09:34:38.678989+00:00","updated_at":"2026-07-05T09:34:38.678989+00:00"}