{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:4G2ZJFDGWQUK6S77IPWJMG7D2X","short_pith_number":"pith:4G2ZJFDG","schema_version":"1.0","canonical_sha256":"e1b5949466b428af4bff43ec961be3d5f142f7ca5581b7ed2b86b2704811a285","source":{"kind":"arxiv","id":"2504.18580","version":1},"attestation_state":"computed","paper":{"title":"Parameter-Efficient Checkpoint Merging via Metrics-Weighted Averaging","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Sehyun Choi, Shi Jie Yu","submitted_at":"2025-04-23T05:11:21Z","abstract_excerpt":"Checkpoint merging is a technique for combining multiple model snapshots into a single superior model, potentially reducing training time for large language models. This paper explores checkpoint merging in the context of parameter-efficient fine-tuning (PEFT), where only small adapter modules (e.g. LoRA) are trained. We propose Metrics-Weighted Averaging (MWA), a simple yet effective method to merge model checkpoints by weighting their parameters according to performance metrics. In particular, we investigate weighting by training loss and by training steps, under the intuition that lower-los"},"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":"2504.18580","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-23T05:11:21Z","cross_cats_sorted":[],"title_canon_sha256":"630723b981950e106cafb978efb78114c0d92efd8027a9ef8834fd1215804abd","abstract_canon_sha256":"b222d72feb6be02c74b735f4fb44b1b360b4ca07255e132a59b0563a54a54566"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:54:17.467723Z","signature_b64":"joIBIbOBd2t4sEeCGlEhFm4SQa0GygCR/aJubXcen7nE4duaOiB1jrZmhoESFkmv4TWmrWWYw5Efe/i613d0BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e1b5949466b428af4bff43ec961be3d5f142f7ca5581b7ed2b86b2704811a285","last_reissued_at":"2026-07-05T10:54:17.467210Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:54:17.467210Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Parameter-Efficient Checkpoint Merging via Metrics-Weighted Averaging","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Sehyun Choi, Shi Jie Yu","submitted_at":"2025-04-23T05:11:21Z","abstract_excerpt":"Checkpoint merging is a technique for combining multiple model snapshots into a single superior model, potentially reducing training time for large language models. This paper explores checkpoint merging in the context of parameter-efficient fine-tuning (PEFT), where only small adapter modules (e.g. LoRA) are trained. We propose Metrics-Weighted Averaging (MWA), a simple yet effective method to merge model checkpoints by weighting their parameters according to performance metrics. In particular, we investigate weighting by training loss and by training steps, under the intuition that lower-los"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.18580","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/2504.18580/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":"2504.18580","created_at":"2026-07-05T10:54:17.467268+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.18580v1","created_at":"2026-07-05T10:54:17.467268+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.18580","created_at":"2026-07-05T10:54:17.467268+00:00"},{"alias_kind":"pith_short_12","alias_value":"4G2ZJFDGWQUK","created_at":"2026-07-05T10:54:17.467268+00:00"},{"alias_kind":"pith_short_16","alias_value":"4G2ZJFDGWQUK6S77","created_at":"2026-07-05T10:54:17.467268+00:00"},{"alias_kind":"pith_short_8","alias_value":"4G2ZJFDG","created_at":"2026-07-05T10:54:17.467268+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/4G2ZJFDGWQUK6S77IPWJMG7D2X","json":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X.json","graph_json":"https://pith.science/api/pith-number/4G2ZJFDGWQUK6S77IPWJMG7D2X/graph.json","events_json":"https://pith.science/api/pith-number/4G2ZJFDGWQUK6S77IPWJMG7D2X/events.json","paper":"https://pith.science/paper/4G2ZJFDG"},"agent_actions":{"view_html":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X","download_json":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X.json","view_paper":"https://pith.science/paper/4G2ZJFDG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.18580&json=true","fetch_graph":"https://pith.science/api/pith-number/4G2ZJFDGWQUK6S77IPWJMG7D2X/graph.json","fetch_events":"https://pith.science/api/pith-number/4G2ZJFDGWQUK6S77IPWJMG7D2X/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X/action/timestamp_anchor","attest_storage":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X/action/storage_attestation","attest_author":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X/action/author_attestation","sign_citation":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X/action/citation_signature","submit_replication":"https://pith.science/pith/4G2ZJFDGWQUK6S77IPWJMG7D2X/action/replication_record"}},"created_at":"2026-07-05T10:54:17.467268+00:00","updated_at":"2026-07-05T10:54:17.467268+00:00"}