{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:JHZZL3AQIZX5TX4GRHGVC4HH4H","short_pith_number":"pith:JHZZL3AQ","schema_version":"1.0","canonical_sha256":"49f395ec10466fd9df8689cd5170e7e1e248f3cca28e1c18126f5608db8f1557","source":{"kind":"arxiv","id":"2312.00968","version":2},"attestation_state":"computed","paper":{"title":"Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Pang, Jialin Wu, Radu Soricut, Xia Hu, Yaqing Wang","submitted_at":"2023-12-01T23:04:27Z","abstract_excerpt":"Large multi-modal models (LMMs) exhibit remarkable performance across numerous tasks. However, generalist LMMs often suffer from performance degradation when tuned over a large collection of tasks. Recent research suggests that Mixture of Experts (MoE) architectures are useful for instruction tuning, but for LMMs of parameter size around O(50-100B), the prohibitive cost of replicating and storing the expert models severely limits the number of experts we can use. We propose Omni-SMoLA, an architecture that uses the Soft MoE approach to (softly) mix many multimodal low rank experts, and avoids "},"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":"2312.00968","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2023-12-01T23:04:27Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"c8468aa07d98db680c6e1a12d0559d019d391e4e572c35238ef69655aa4f7b89","abstract_canon_sha256":"b4aac12cf9cff5e7a66bcbdeeec54091e616ad8100319feb76b2e9cd49ed1679"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:03:46.372416Z","signature_b64":"K5DlYajMu50ARHm5tf0iLQ2uJhs+zf7pJL5vu0AgMPaauwrgxBY1q2gbUThIOW1d8uHPoFD2cvdrkvoi2mDLCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"49f395ec10466fd9df8689cd5170e7e1e248f3cca28e1c18126f5608db8f1557","last_reissued_at":"2026-07-05T08:03:46.371947Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:03:46.371947Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Omni-SMoLA: Boosting Generalist Multimodal Models with Soft Mixture of Low-rank Experts","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Bo Pang, Jialin Wu, Radu Soricut, Xia Hu, Yaqing Wang","submitted_at":"2023-12-01T23:04:27Z","abstract_excerpt":"Large multi-modal models (LMMs) exhibit remarkable performance across numerous tasks. However, generalist LMMs often suffer from performance degradation when tuned over a large collection of tasks. Recent research suggests that Mixture of Experts (MoE) architectures are useful for instruction tuning, but for LMMs of parameter size around O(50-100B), the prohibitive cost of replicating and storing the expert models severely limits the number of experts we can use. We propose Omni-SMoLA, an architecture that uses the Soft MoE approach to (softly) mix many multimodal low rank experts, and avoids "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2312.00968","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/2312.00968/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":"2312.00968","created_at":"2026-07-05T08:03:46.372008+00:00"},{"alias_kind":"arxiv_version","alias_value":"2312.00968v2","created_at":"2026-07-05T08:03:46.372008+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2312.00968","created_at":"2026-07-05T08:03:46.372008+00:00"},{"alias_kind":"pith_short_12","alias_value":"JHZZL3AQIZX5","created_at":"2026-07-05T08:03:46.372008+00:00"},{"alias_kind":"pith_short_16","alias_value":"JHZZL3AQIZX5TX4G","created_at":"2026-07-05T08:03:46.372008+00:00"},{"alias_kind":"pith_short_8","alias_value":"JHZZL3AQ","created_at":"2026-07-05T08:03:46.372008+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2608.00574","citing_title":"Relax Within, Balance Across: Geometry-Guided Load Balancing for Vision-Language Mixture-of-Experts","ref_index":11,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H","json":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H.json","graph_json":"https://pith.science/api/pith-number/JHZZL3AQIZX5TX4GRHGVC4HH4H/graph.json","events_json":"https://pith.science/api/pith-number/JHZZL3AQIZX5TX4GRHGVC4HH4H/events.json","paper":"https://pith.science/paper/JHZZL3AQ"},"agent_actions":{"view_html":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H","download_json":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H.json","view_paper":"https://pith.science/paper/JHZZL3AQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2312.00968&json=true","fetch_graph":"https://pith.science/api/pith-number/JHZZL3AQIZX5TX4GRHGVC4HH4H/graph.json","fetch_events":"https://pith.science/api/pith-number/JHZZL3AQIZX5TX4GRHGVC4HH4H/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H/action/storage_attestation","attest_author":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H/action/author_attestation","sign_citation":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H/action/citation_signature","submit_replication":"https://pith.science/pith/JHZZL3AQIZX5TX4GRHGVC4HH4H/action/replication_record"}},"created_at":"2026-07-05T08:03:46.372008+00:00","updated_at":"2026-07-05T08:03:46.372008+00:00"}