{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:UP7OHPXHZ4NNATKNZIQ3TIMQL7","short_pith_number":"pith:UP7OHPXH","schema_version":"1.0","canonical_sha256":"a3fee3bee7cf1ad04d4dca21b9a1905fd47c3e1d601baaba6e23c4450c737e08","source":{"kind":"arxiv","id":"2406.16330","version":2},"attestation_state":"computed","paper":{"title":"Pruning via Merging: Compressing LLMs via Manifold Alignment Based Layer Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Li, Cunhang Fan, Deyuan Liu, Dianbo Sui, Dianhui Chu, Fangying Rong, Hairu Wang, Qingbin Liu, Xi Chen, YanChao Hao, Zecheng Wang, Zhanyue Qin, Zhao Lv, Zhao Yang, Zhiying Tu","submitted_at":"2024-06-24T05:57:55Z","abstract_excerpt":"While large language models (LLMs) excel in many domains, their complexity and scale challenge deployment in resource-limited environments. Current compression techniques, such as parameter pruning, often fail to effectively utilize the knowledge from pruned parameters. To address these challenges, we propose Manifold-Based Knowledge Alignment and Layer Merging Compression (MKA), a novel approach that uses manifold learning and the Normalized Pairwise Information Bottleneck (NPIB) measure to merge similar layers, reducing model size while preserving essential performance. We evaluate MKA on mu"},"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":"2406.16330","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-24T05:57:55Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"bb3a217571e8f0ac44e0bc52cd75f4e6e3b6bb050e4b1a7c00ef233257b42226","abstract_canon_sha256":"5c09426b7c7dd1f2d4c4248ae7a4b48c503c374f9fa99f1609d50e7d5a3ec89f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:04:17.939013Z","signature_b64":"L7KEDmu1p3kamsLkuz6wKAsSLCOeV2u/EZE+7baZCexEq5y+zb/Pdbu3HHq4MgITYW7GCcrlfW65U0gkI/KGDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a3fee3bee7cf1ad04d4dca21b9a1905fd47c3e1d601baaba6e23c4450c737e08","last_reissued_at":"2026-07-05T11:04:17.938476Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:04:17.938476Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Pruning via Merging: Compressing LLMs via Manifold Alignment Based Layer Merging","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bo Li, Cunhang Fan, Deyuan Liu, Dianbo Sui, Dianhui Chu, Fangying Rong, Hairu Wang, Qingbin Liu, Xi Chen, YanChao Hao, Zecheng Wang, Zhanyue Qin, Zhao Lv, Zhao Yang, Zhiying Tu","submitted_at":"2024-06-24T05:57:55Z","abstract_excerpt":"While large language models (LLMs) excel in many domains, their complexity and scale challenge deployment in resource-limited environments. Current compression techniques, such as parameter pruning, often fail to effectively utilize the knowledge from pruned parameters. To address these challenges, we propose Manifold-Based Knowledge Alignment and Layer Merging Compression (MKA), a novel approach that uses manifold learning and the Normalized Pairwise Information Bottleneck (NPIB) measure to merge similar layers, reducing model size while preserving essential performance. We evaluate MKA on mu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.16330","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/2406.16330/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":"2406.16330","created_at":"2026-07-05T11:04:17.938537+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.16330v2","created_at":"2026-07-05T11:04:17.938537+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.16330","created_at":"2026-07-05T11:04:17.938537+00:00"},{"alias_kind":"pith_short_12","alias_value":"UP7OHPXHZ4NN","created_at":"2026-07-05T11:04:17.938537+00:00"},{"alias_kind":"pith_short_16","alias_value":"UP7OHPXHZ4NNATKN","created_at":"2026-07-05T11:04:17.938537+00:00"},{"alias_kind":"pith_short_8","alias_value":"UP7OHPXH","created_at":"2026-07-05T11:04:17.938537+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.12358","citing_title":"Why and When Visual Token Pruning Fails? A Study on Relevant Visual Information Shift in MLLMs Decoding","ref_index":28,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07271","citing_title":"Understanding Performance Collapse in Layer-Pruned Large Language Models via Decision Representation Transitions","ref_index":57,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7","json":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7.json","graph_json":"https://pith.science/api/pith-number/UP7OHPXHZ4NNATKNZIQ3TIMQL7/graph.json","events_json":"https://pith.science/api/pith-number/UP7OHPXHZ4NNATKNZIQ3TIMQL7/events.json","paper":"https://pith.science/paper/UP7OHPXH"},"agent_actions":{"view_html":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7","download_json":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7.json","view_paper":"https://pith.science/paper/UP7OHPXH","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.16330&json=true","fetch_graph":"https://pith.science/api/pith-number/UP7OHPXHZ4NNATKNZIQ3TIMQL7/graph.json","fetch_events":"https://pith.science/api/pith-number/UP7OHPXHZ4NNATKNZIQ3TIMQL7/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7/action/storage_attestation","attest_author":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7/action/author_attestation","sign_citation":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7/action/citation_signature","submit_replication":"https://pith.science/pith/UP7OHPXHZ4NNATKNZIQ3TIMQL7/action/replication_record"}},"created_at":"2026-07-05T11:04:17.938537+00:00","updated_at":"2026-07-05T11:04:17.938537+00:00"}