{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:3OQ3RTWGYWQAGZHWMFLCNXYYFU","short_pith_number":"pith:3OQ3RTWG","schema_version":"1.0","canonical_sha256":"dba1b8cec6c5a00364f6615626df182d2589635a9b47d26db1bff631155e5436","source":{"kind":"arxiv","id":"2507.00029","version":2},"attestation_state":"computed","paper":{"title":"LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention Routing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hang Zhou, Junqing Yu, Wei Yang, Wenbing Li, Yunyao Zhang, Zikai Song","submitted_at":"2025-06-17T14:58:54Z","abstract_excerpt":"Recent attempts to combine low-rank adaptation (LoRA) with mixture-of-experts (MoE) for multi-task adaptation of Large Language Models (LLMs) often replace whole attention/FFN layers with switch experts or append parallel expert branches, undermining parameter efficiency and limiting task specialization. We introduce LoRA-Mixer, a modular MoE framework that routes task-specific LoRA experts into the core projection matrices of the attention module, namely input and output linear layers, rather than primarily targeting FFN blocks. The design delivers fine-grained token-level specialization by f"},"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":"2507.00029","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-17T14:58:54Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e8d0a9dae3423f38d4a1d2c785498996294e97a7dc5974f91a08daffba5af6c3","abstract_canon_sha256":"4dd94841bb2da2eeab74c8bbd2fa692e43f3c2b49b125204d1b9a20b8fb3bb79"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:09:35.324358Z","signature_b64":"2zPqePfYElF1tiDowQ1yKJUZXyLFXvrqsFEGWqzZ8M00hV0nRZ6+RAw5hHhrGicLrE0+5L/FJ9KrIbItwHH7Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dba1b8cec6c5a00364f6615626df182d2589635a9b47d26db1bff631155e5436","last_reissued_at":"2026-05-18T03:09:35.323714Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:09:35.323714Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LoRA-Mixer: Coordinate Modular LoRA Experts Through Serial Attention Routing","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Hang Zhou, Junqing Yu, Wei Yang, Wenbing Li, Yunyao Zhang, Zikai Song","submitted_at":"2025-06-17T14:58:54Z","abstract_excerpt":"Recent attempts to combine low-rank adaptation (LoRA) with mixture-of-experts (MoE) for multi-task adaptation of Large Language Models (LLMs) often replace whole attention/FFN layers with switch experts or append parallel expert branches, undermining parameter efficiency and limiting task specialization. We introduce LoRA-Mixer, a modular MoE framework that routes task-specific LoRA experts into the core projection matrices of the attention module, namely input and output linear layers, rather than primarily targeting FFN blocks. The design delivers fine-grained token-level specialization by f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.00029","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":""},"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":"2507.00029","created_at":"2026-05-18T03:09:35.323873+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.00029v2","created_at":"2026-05-18T03:09:35.323873+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.00029","created_at":"2026-05-18T03:09:35.323873+00:00"},{"alias_kind":"pith_short_12","alias_value":"3OQ3RTWGYWQA","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"3OQ3RTWGYWQAGZHW","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"3OQ3RTWG","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":6,"sample":[{"citing_arxiv_id":"2604.26707","citing_title":"CurEvo: Curriculum-Guided Self-Evolution for Video Understanding","ref_index":40,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26353","citing_title":"GateMOT: Q-Gated Attention for Dense Object Tracking","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2604.26252","citing_title":"OmniTrend: Content-Context Modeling for Scalable Social Popularity Prediction","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25614","citing_title":"HotComment: A Benchmark for Evaluating Popularity of Online Comments","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2604.20311","citing_title":"Seeing Further and Wider: Joint Spatio-Temporal Enlargement for Micro-Video Popularity Prediction","ref_index":39,"is_internal_anchor":true},{"citing_arxiv_id":"2604.06600","citing_title":"IntervenSim: Intervention-Aware Social Network Simulation for Opinion Dynamics","ref_index":43,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU","json":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU.json","graph_json":"https://pith.science/api/pith-number/3OQ3RTWGYWQAGZHWMFLCNXYYFU/graph.json","events_json":"https://pith.science/api/pith-number/3OQ3RTWGYWQAGZHWMFLCNXYYFU/events.json","paper":"https://pith.science/paper/3OQ3RTWG"},"agent_actions":{"view_html":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU","download_json":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU.json","view_paper":"https://pith.science/paper/3OQ3RTWG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.00029&json=true","fetch_graph":"https://pith.science/api/pith-number/3OQ3RTWGYWQAGZHWMFLCNXYYFU/graph.json","fetch_events":"https://pith.science/api/pith-number/3OQ3RTWGYWQAGZHWMFLCNXYYFU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU/action/storage_attestation","attest_author":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU/action/author_attestation","sign_citation":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU/action/citation_signature","submit_replication":"https://pith.science/pith/3OQ3RTWGYWQAGZHWMFLCNXYYFU/action/replication_record"}},"created_at":"2026-05-18T03:09:35.323873+00:00","updated_at":"2026-05-18T03:09:35.323873+00:00"}