{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:JLSYJDGEG5QQI5PRL233TIZD3U","short_pith_number":"pith:JLSYJDGE","schema_version":"1.0","canonical_sha256":"4ae5848cc437610475f15eb7b9a323dd0560a812dfee8703c6417eb79a540f92","source":{"kind":"arxiv","id":"2606.03391","version":1},"attestation_state":"computed","paper":{"title":"When Model Merging Breaks Routing: Training-Free Calibration for MoE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Canbin Huang, Jianfei Zhang, Jingang Wang, Qifan Wang, Tianyuan Shi, Xiaojun Quan","submitted_at":"2026-06-02T09:33:33Z","abstract_excerpt":"Model merging has emerged as a cost-effective approach for consolidating the capabilities of multiple LLMs without retraining. However, existing merging techniques, largely based on linear parameter arithmetic or optimization, struggle when applied to Mixture-of-Experts (MoE) architectures. We identify a critical failure mode in MoE merging, termed routing breakdown, in which the merged router fails to dispatch tokens to suitable experts. Routing breakdown stems from the sensitivity of the non-linear softmax and discrete Top-k routing mechanisms to parameter perturbations from merging, a sensi"},"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":"2606.03391","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-02T09:33:33Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"db456afabcb306e3c6b4808056ecce363ec7f338233d82978fc4d514caf5b9b7","abstract_canon_sha256":"654e5b160b8c8ad88f348c4a47eccf1370f9c07e1dad17b87b48533ae0bd8d9f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T01:05:56.675438Z","signature_b64":"2H76FHD6TUlAEWin7bM/nM4pLJluVTcdNG0TmTLOBi4Y+t8I1OUSv11h2faB/2fiQCAl6isKvjVs+B1fSqxaAQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4ae5848cc437610475f15eb7b9a323dd0560a812dfee8703c6417eb79a540f92","last_reissued_at":"2026-06-03T01:05:56.675089Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T01:05:56.675089Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When Model Merging Breaks Routing: Training-Free Calibration for MoE","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Canbin Huang, Jianfei Zhang, Jingang Wang, Qifan Wang, Tianyuan Shi, Xiaojun Quan","submitted_at":"2026-06-02T09:33:33Z","abstract_excerpt":"Model merging has emerged as a cost-effective approach for consolidating the capabilities of multiple LLMs without retraining. However, existing merging techniques, largely based on linear parameter arithmetic or optimization, struggle when applied to Mixture-of-Experts (MoE) architectures. We identify a critical failure mode in MoE merging, termed routing breakdown, in which the merged router fails to dispatch tokens to suitable experts. Routing breakdown stems from the sensitivity of the non-linear softmax and discrete Top-k routing mechanisms to parameter perturbations from merging, a sensi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.03391","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/2606.03391/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":"2606.03391","created_at":"2026-06-03T01:05:56.675146+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.03391v1","created_at":"2026-06-03T01:05:56.675146+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.03391","created_at":"2026-06-03T01:05:56.675146+00:00"},{"alias_kind":"pith_short_12","alias_value":"JLSYJDGEG5QQ","created_at":"2026-06-03T01:05:56.675146+00:00"},{"alias_kind":"pith_short_16","alias_value":"JLSYJDGEG5QQI5PR","created_at":"2026-06-03T01:05:56.675146+00:00"},{"alias_kind":"pith_short_8","alias_value":"JLSYJDGE","created_at":"2026-06-03T01:05:56.675146+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/JLSYJDGEG5QQI5PRL233TIZD3U","json":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U.json","graph_json":"https://pith.science/api/pith-number/JLSYJDGEG5QQI5PRL233TIZD3U/graph.json","events_json":"https://pith.science/api/pith-number/JLSYJDGEG5QQI5PRL233TIZD3U/events.json","paper":"https://pith.science/paper/JLSYJDGE"},"agent_actions":{"view_html":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U","download_json":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U.json","view_paper":"https://pith.science/paper/JLSYJDGE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.03391&json=true","fetch_graph":"https://pith.science/api/pith-number/JLSYJDGEG5QQI5PRL233TIZD3U/graph.json","fetch_events":"https://pith.science/api/pith-number/JLSYJDGEG5QQI5PRL233TIZD3U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U/action/storage_attestation","attest_author":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U/action/author_attestation","sign_citation":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U/action/citation_signature","submit_replication":"https://pith.science/pith/JLSYJDGEG5QQI5PRL233TIZD3U/action/replication_record"}},"created_at":"2026-06-03T01:05:56.675146+00:00","updated_at":"2026-06-03T01:05:56.675146+00:00"}