{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FM6KO3CVRVPLWEIIGANZZHGPSK","short_pith_number":"pith:FM6KO3CV","schema_version":"1.0","canonical_sha256":"2b3ca76c558d5ebb1108301b9c9ccf929e443b99a9c7b730a91581e2d0f576e0","source":{"kind":"arxiv","id":"2505.24593","version":2},"attestation_state":"computed","paper":{"title":"Decoding Knowledge Attribution in Mixture-of-Experts: A Framework of Basic-Refinement Collaboration and Efficiency Analysis","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bo Wang, Jia Liu, Junzhuo Li, Peijie Jiang, Xiuze Zhou, Xuming Hu","submitted_at":"2025-05-30T13:40:51Z","abstract_excerpt":"The interpretability of Mixture-of-Experts (MoE) models, especially those with heterogeneous designs, remains underexplored. Existing attribution methods for dense models fail to capture dynamic routing-expert interactions in sparse MoE architectures. To address this issue, we propose a cross-level attribution algorithm to analyze sparse MoE architectures (Qwen 1.5-MoE, OLMoE, Mixtral-8x7B) against dense models (Qwen 1.5-7B, Llama-7B, Mistral-7B). Results show MoE models achieve 37% higher per-layer efficiency via a \"mid-activation, late-amplification\" pattern: early layers screen experts, whi"},"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":"2505.24593","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-05-30T13:40:51Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"73dc2f525c849c600957a05cfedabc3f3505c1b67bff0db9dd36fe3466aed7f7","abstract_canon_sha256":"f195f586c7b46335cfdbf9af0793d750b309edc635fb5c0284e5dc97d8184d24"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:19:27.573722Z","signature_b64":"8SG3DO3QcP2rPy5JynYN4WDFgmAEI1L2Sn+xAN/3o4GYAAquf5SgSwuDRFqbqAJ0hUHWSaEZ0o69LmHgiGRUCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2b3ca76c558d5ebb1108301b9c9ccf929e443b99a9c7b730a91581e2d0f576e0","last_reissued_at":"2026-07-05T11:19:27.573220Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:19:27.573220Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Decoding Knowledge Attribution in Mixture-of-Experts: A Framework of Basic-Refinement Collaboration and Efficiency Analysis","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Bo Wang, Jia Liu, Junzhuo Li, Peijie Jiang, Xiuze Zhou, Xuming Hu","submitted_at":"2025-05-30T13:40:51Z","abstract_excerpt":"The interpretability of Mixture-of-Experts (MoE) models, especially those with heterogeneous designs, remains underexplored. Existing attribution methods for dense models fail to capture dynamic routing-expert interactions in sparse MoE architectures. To address this issue, we propose a cross-level attribution algorithm to analyze sparse MoE architectures (Qwen 1.5-MoE, OLMoE, Mixtral-8x7B) against dense models (Qwen 1.5-7B, Llama-7B, Mistral-7B). Results show MoE models achieve 37% higher per-layer efficiency via a \"mid-activation, late-amplification\" pattern: early layers screen experts, whi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.24593","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/2505.24593/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":"2505.24593","created_at":"2026-07-05T11:19:27.573279+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.24593v2","created_at":"2026-07-05T11:19:27.573279+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.24593","created_at":"2026-07-05T11:19:27.573279+00:00"},{"alias_kind":"pith_short_12","alias_value":"FM6KO3CVRVPL","created_at":"2026-07-05T11:19:27.573279+00:00"},{"alias_kind":"pith_short_16","alias_value":"FM6KO3CVRVPLWEII","created_at":"2026-07-05T11:19:27.573279+00:00"},{"alias_kind":"pith_short_8","alias_value":"FM6KO3CV","created_at":"2026-07-05T11:19:27.573279+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2508.02312","citing_title":"A Survey on Data Security in Large Language Models","ref_index":36,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK","json":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK.json","graph_json":"https://pith.science/api/pith-number/FM6KO3CVRVPLWEIIGANZZHGPSK/graph.json","events_json":"https://pith.science/api/pith-number/FM6KO3CVRVPLWEIIGANZZHGPSK/events.json","paper":"https://pith.science/paper/FM6KO3CV"},"agent_actions":{"view_html":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK","download_json":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK.json","view_paper":"https://pith.science/paper/FM6KO3CV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.24593&json=true","fetch_graph":"https://pith.science/api/pith-number/FM6KO3CVRVPLWEIIGANZZHGPSK/graph.json","fetch_events":"https://pith.science/api/pith-number/FM6KO3CVRVPLWEIIGANZZHGPSK/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK/action/storage_attestation","attest_author":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK/action/author_attestation","sign_citation":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK/action/citation_signature","submit_replication":"https://pith.science/pith/FM6KO3CVRVPLWEIIGANZZHGPSK/action/replication_record"}},"created_at":"2026-07-05T11:19:27.573279+00:00","updated_at":"2026-07-05T11:19:27.573279+00:00"}