{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:W57O62RG3ETKLQZN3RFMEXTMJX","short_pith_number":"pith:W57O62RG","schema_version":"1.0","canonical_sha256":"b77eef6a26d926a5c32ddc4ac25e6c4de4c1b7d551c8cd5e9197ee179e4dc460","source":{"kind":"arxiv","id":"2608.05000","version":1},"attestation_state":"computed","paper":{"title":"Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"David Fan, Filippos Kokkinos, Junlin Han, Mike Lewis, Minghao Chen, Philip Torr, Shengbang Tong","submitted_at":"2026-08-05T16:09:25Z","abstract_excerpt":"Vision offers a critical axis for advancing foundation models, driving a shift towards natively unified multimodal pretraining. Despite this momentum, the design space and the fundamental mechanisms of how modalities interact during unified training remain underexplored. We provide empirical clarity through a systematic exploration of multimodal pretraining. Our controlled experiments on both synthetic and large-scale real-world datasets yield four key insights into the physics of multimodal pretraining: (i) Knowledge Flow: We disentangle how language, visual understanding, and visual generati"},"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":"2608.05000","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2026-08-05T16:09:25Z","cross_cats_sorted":["cs.LG","cs.MM"],"title_canon_sha256":"1a8291b49ebfe6b01a3010ae9d170b957421ee6d0b017c7ed8fdbd2db502dcf7","abstract_canon_sha256":"177fee1269536ccc240d5cc64a2374aa54731cef1645ac7bcc5d74c4a5c31232"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-06T01:48:01.704197Z","signature_b64":"4iteRx5hl3X9qVaomchpOsoJc8R+r+qD90EzZi9J7QIRjVk7chT6sLyRubYcxFPc+pz9zG3+rjZ/giB+blqcBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b77eef6a26d926a5c32ddc4ac25e6c4de4c1b7d551c8cd5e9197ee179e4dc460","last_reissued_at":"2026-08-06T01:48:01.702767Z","signature_status":"signed_v1","first_computed_at":"2026-08-06T01:48:01.702767Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.MM"],"primary_cat":"cs.CV","authors_text":"David Fan, Filippos Kokkinos, Junlin Han, Mike Lewis, Minghao Chen, Philip Torr, Shengbang Tong","submitted_at":"2026-08-05T16:09:25Z","abstract_excerpt":"Vision offers a critical axis for advancing foundation models, driving a shift towards natively unified multimodal pretraining. Despite this momentum, the design space and the fundamental mechanisms of how modalities interact during unified training remain underexplored. We provide empirical clarity through a systematic exploration of multimodal pretraining. Our controlled experiments on both synthetic and large-scale real-world datasets yield four key insights into the physics of multimodal pretraining: (i) Knowledge Flow: We disentangle how language, visual understanding, and visual generati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.05000","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/2608.05000/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":"2608.05000","created_at":"2026-08-06T01:48:01.704578+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.05000v1","created_at":"2026-08-06T01:48:01.704578+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.05000","created_at":"2026-08-06T01:48:01.704578+00:00"},{"alias_kind":"pith_short_12","alias_value":"W57O62RG3ETK","created_at":"2026-08-06T01:48:01.704578+00:00"},{"alias_kind":"pith_short_16","alias_value":"W57O62RG3ETKLQZN","created_at":"2026-08-06T01:48:01.704578+00:00"},{"alias_kind":"pith_short_8","alias_value":"W57O62RG","created_at":"2026-08-06T01:48:01.704578+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/W57O62RG3ETKLQZN3RFMEXTMJX","json":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX.json","graph_json":"https://pith.science/api/pith-number/W57O62RG3ETKLQZN3RFMEXTMJX/graph.json","events_json":"https://pith.science/api/pith-number/W57O62RG3ETKLQZN3RFMEXTMJX/events.json","paper":"https://pith.science/paper/W57O62RG"},"agent_actions":{"view_html":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX","download_json":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX.json","view_paper":"https://pith.science/paper/W57O62RG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.05000&json=true","fetch_graph":"https://pith.science/api/pith-number/W57O62RG3ETKLQZN3RFMEXTMJX/graph.json","fetch_events":"https://pith.science/api/pith-number/W57O62RG3ETKLQZN3RFMEXTMJX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX/action/storage_attestation","attest_author":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX/action/author_attestation","sign_citation":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX/action/citation_signature","submit_replication":"https://pith.science/pith/W57O62RG3ETKLQZN3RFMEXTMJX/action/replication_record"}},"created_at":"2026-08-06T01:48:01.704578+00:00","updated_at":"2026-08-06T01:48:01.704578+00:00"}