{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:WFBFMNVWK22GUF62LDPTQJZFET","short_pith_number":"pith:WFBFMNVW","schema_version":"1.0","canonical_sha256":"b1425636b656b46a17da58df38272524fb222ce813399d8e99c09442f7a1a447","source":{"kind":"arxiv","id":"2504.20996","version":1},"attestation_state":"computed","paper":{"title":"X-Fusion: Introducing New Modality to Frozen Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Abhishek Tandon, Bolei Zhou, Eli Shechtman, Krishna Kumar Singh, Sicheng Mo, Siddharth Srinivasan Iyer, Thao Nguyen, Xun Huang, Yijun Li, Yong Jae Lee, Yuchen Liu, Yuheng Li","submitted_at":"2025-04-29T17:59:45Z","abstract_excerpt":"We propose X-Fusion, a framework that extends pretrained Large Language Models (LLMs) for multimodal tasks while preserving their language capabilities. X-Fusion employs a dual-tower design with modality-specific weights, keeping the LLM's parameters frozen while integrating vision-specific information for both understanding and generation. Our experiments demonstrate that X-Fusion consistently outperforms alternative architectures on both image-to-text and text-to-image tasks. We find that incorporating understanding-focused data improves generation quality, reducing image data noise enhances"},"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":"2504.20996","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-04-29T17:59:45Z","cross_cats_sorted":[],"title_canon_sha256":"927d069b323ee79923b722a3e1f9d8a778a51bb76af233d1d43f863f1ff14bf6","abstract_canon_sha256":"771c2e4f56d62704cd6c7057ac8a23f355b39eff511a749f71791ec53f8ceb71"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:55:52.749457Z","signature_b64":"lUbBhbypvs1AKITFgIk4VTGgyXm0hGA9SWdIl/XQRLMwtb3z8bHRdg26jO1tOpZOWYvJ6rh5aC4ZHDdDyNPQBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b1425636b656b46a17da58df38272524fb222ce813399d8e99c09442f7a1a447","last_reissued_at":"2026-07-05T10:55:52.748995Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:55:52.748995Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"X-Fusion: Introducing New Modality to Frozen Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Abhishek Tandon, Bolei Zhou, Eli Shechtman, Krishna Kumar Singh, Sicheng Mo, Siddharth Srinivasan Iyer, Thao Nguyen, Xun Huang, Yijun Li, Yong Jae Lee, Yuchen Liu, Yuheng Li","submitted_at":"2025-04-29T17:59:45Z","abstract_excerpt":"We propose X-Fusion, a framework that extends pretrained Large Language Models (LLMs) for multimodal tasks while preserving their language capabilities. X-Fusion employs a dual-tower design with modality-specific weights, keeping the LLM's parameters frozen while integrating vision-specific information for both understanding and generation. Our experiments demonstrate that X-Fusion consistently outperforms alternative architectures on both image-to-text and text-to-image tasks. We find that incorporating understanding-focused data improves generation quality, reducing image data noise enhances"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.20996","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/2504.20996/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":"2504.20996","created_at":"2026-07-05T10:55:52.749041+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.20996v1","created_at":"2026-07-05T10:55:52.749041+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.20996","created_at":"2026-07-05T10:55:52.749041+00:00"},{"alias_kind":"pith_short_12","alias_value":"WFBFMNVWK22G","created_at":"2026-07-05T10:55:52.749041+00:00"},{"alias_kind":"pith_short_16","alias_value":"WFBFMNVWK22GUF62","created_at":"2026-07-05T10:55:52.749041+00:00"},{"alias_kind":"pith_short_8","alias_value":"WFBFMNVW","created_at":"2026-07-05T10:55:52.749041+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.08009","citing_title":"Self Forcing: Bridging the Train-Test Gap in Autoregressive Video Diffusion","ref_index":59,"is_internal_anchor":false},{"citing_arxiv_id":"2604.07209","citing_title":"INSPATIO-WORLD: A Real-Time 4D World Simulator via Spatiotemporal Autoregressive Modeling","ref_index":63,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET","json":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET.json","graph_json":"https://pith.science/api/pith-number/WFBFMNVWK22GUF62LDPTQJZFET/graph.json","events_json":"https://pith.science/api/pith-number/WFBFMNVWK22GUF62LDPTQJZFET/events.json","paper":"https://pith.science/paper/WFBFMNVW"},"agent_actions":{"view_html":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET","download_json":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET.json","view_paper":"https://pith.science/paper/WFBFMNVW","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.20996&json=true","fetch_graph":"https://pith.science/api/pith-number/WFBFMNVWK22GUF62LDPTQJZFET/graph.json","fetch_events":"https://pith.science/api/pith-number/WFBFMNVWK22GUF62LDPTQJZFET/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET/action/timestamp_anchor","attest_storage":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET/action/storage_attestation","attest_author":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET/action/author_attestation","sign_citation":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET/action/citation_signature","submit_replication":"https://pith.science/pith/WFBFMNVWK22GUF62LDPTQJZFET/action/replication_record"}},"created_at":"2026-07-05T10:55:52.749041+00:00","updated_at":"2026-07-05T10:55:52.749041+00:00"}