{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:TH475T5TXVBBTZNMRHFEBAIZDJ","short_pith_number":"pith:TH475T5T","schema_version":"1.0","canonical_sha256":"99f9fecfb3bd4219e5ac89ca4081191a6f6266ad0b0cdd4d987c016d8d450dd8","source":{"kind":"arxiv","id":"2505.13893","version":2},"attestation_state":"computed","paper":{"title":"InfiGFusion: Graph-on-Logits Distillation via Efficient Gromov-Wasserstein for Model Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fei Wu, Hongxia Yang, Qi Zhou, Yanggan Gu, Yiming Zhang, Yuanyi Wang, Zhaoyi Yan","submitted_at":"2025-05-20T03:55:35Z","abstract_excerpt":"Recent advances in large language models (LLMs) have intensified efforts to fuse heterogeneous open-source models into a unified system that inherits their complementary strengths. Existing logit-based fusion methods maintain inference efficiency but treat vocabulary dimensions independently, overlooking semantic dependencies encoded by cross-dimension interactions. These dependencies reflect how token types interact under a model's internal reasoning and are essential for aligning models with diverse generation behaviors. To explicitly model these dependencies, we propose \\textbf{InfiGFusion}"},"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":"2505.13893","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-20T03:55:35Z","cross_cats_sorted":[],"title_canon_sha256":"4719a30cabb35ba5d723dda59523b56c4f6968c97515b1c6157cde9bd3f1b1d2","abstract_canon_sha256":"6038b48d95e3a4625107223700a42e347169d0753a786cd9d8f7cd812a75f5cd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-25T02:01:04.349106Z","signature_b64":"MF+SbLW6TNPtQo/QhKNyneRWFq8SLYeV1LJde7J7vyDEGHPKFye4pePdCzwIDGXQ6GWCnkJhygsvmm1u/EzECA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"99f9fecfb3bd4219e5ac89ca4081191a6f6266ad0b0cdd4d987c016d8d450dd8","last_reissued_at":"2026-05-25T02:01:04.348345Z","signature_status":"signed_v1","first_computed_at":"2026-05-25T02:01:04.348345Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"InfiGFusion: Graph-on-Logits Distillation via Efficient Gromov-Wasserstein for Model Fusion","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Fei Wu, Hongxia Yang, Qi Zhou, Yanggan Gu, Yiming Zhang, Yuanyi Wang, Zhaoyi Yan","submitted_at":"2025-05-20T03:55:35Z","abstract_excerpt":"Recent advances in large language models (LLMs) have intensified efforts to fuse heterogeneous open-source models into a unified system that inherits their complementary strengths. Existing logit-based fusion methods maintain inference efficiency but treat vocabulary dimensions independently, overlooking semantic dependencies encoded by cross-dimension interactions. These dependencies reflect how token types interact under a model's internal reasoning and are essential for aligning models with diverse generation behaviors. To explicitly model these dependencies, we propose \\textbf{InfiGFusion}"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.13893","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/2505.13893/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":"2505.13893","created_at":"2026-05-25T02:01:04.348461+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.13893v2","created_at":"2026-05-25T02:01:04.348461+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.13893","created_at":"2026-05-25T02:01:04.348461+00:00"},{"alias_kind":"pith_short_12","alias_value":"TH475T5TXVBB","created_at":"2026-05-25T02:01:04.348461+00:00"},{"alias_kind":"pith_short_16","alias_value":"TH475T5TXVBBTZNM","created_at":"2026-05-25T02:01:04.348461+00:00"},{"alias_kind":"pith_short_8","alias_value":"TH475T5T","created_at":"2026-05-25T02:01:04.348461+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":4,"sample":[{"citing_arxiv_id":"2605.16882","citing_title":"E-PMQ: Expert-Guided Post-Merge Quantization with Merged-Weight Anchoring","ref_index":9,"is_internal_anchor":true},{"citing_arxiv_id":"2605.14546","citing_title":"Discovering Physical Directions in Weight Space: Composing Neural PDE Experts","ref_index":43,"is_internal_anchor":true},{"citing_arxiv_id":"2605.13030","citing_title":"FeatCal: Feature Calibration for Post-Merging Models","ref_index":62,"is_internal_anchor":true},{"citing_arxiv_id":"2605.09608","citing_title":"Geometry Conflict: Explaining and Controlling Forgetting in LLM Continual Post-Training","ref_index":68,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ","json":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ.json","graph_json":"https://pith.science/api/pith-number/TH475T5TXVBBTZNMRHFEBAIZDJ/graph.json","events_json":"https://pith.science/api/pith-number/TH475T5TXVBBTZNMRHFEBAIZDJ/events.json","paper":"https://pith.science/paper/TH475T5T"},"agent_actions":{"view_html":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ","download_json":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ.json","view_paper":"https://pith.science/paper/TH475T5T","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.13893&json=true","fetch_graph":"https://pith.science/api/pith-number/TH475T5TXVBBTZNMRHFEBAIZDJ/graph.json","fetch_events":"https://pith.science/api/pith-number/TH475T5TXVBBTZNMRHFEBAIZDJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ/action/storage_attestation","attest_author":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ/action/author_attestation","sign_citation":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ/action/citation_signature","submit_replication":"https://pith.science/pith/TH475T5TXVBBTZNMRHFEBAIZDJ/action/replication_record"}},"created_at":"2026-05-25T02:01:04.348461+00:00","updated_at":"2026-05-25T02:01:04.348461+00:00"}