{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BEZMONTNBHOM5NNU7MQDFB6YZE","short_pith_number":"pith:BEZMONTN","schema_version":"1.0","canonical_sha256":"0932c7366d09dcceb5b4fb203287d8c90e8ce38333c72a10b341191509eda14e","source":{"kind":"arxiv","id":"2502.10436","version":4},"attestation_state":"computed","paper":{"title":"MERGE$^3$: Efficient Evolutionary Merging on Consumer-grade GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.NE","authors_text":"Adrian Robert Minut, Andrea Santilli, Donato Crisostomi, Emanuele Rodol\\`a, Tommaso Mencattini","submitted_at":"2025-02-09T14:24:16Z","abstract_excerpt":"Evolutionary model merging enables the creation of high-performing multi-task models but remains computationally prohibitive for consumer hardware. We introduce MERGE$^3$, an efficient framework that makes evolutionary merging feasible on a single GPU by reducing fitness computation costs 50$\\times$ while preserving performance. MERGE$^3$ achieves this by Extracting a reduced dataset for evaluation, Estimating model abilities using Item Response Theory (IRT), and Evolving optimal merges via IRT-based performance estimators. Our method enables state-of-the-art multilingual and cross-lingual mer"},"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":"2502.10436","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.NE","submitted_at":"2025-02-09T14:24:16Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"85bc8c9417329723b1f40723184bfb301c6c2f8e9cdd8d1646a908d5c4fa4203","abstract_canon_sha256":"068c6fd45075a6bb206cb334973be2bdce30d33db226f954ca16e18511145ec4"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:00:42.213134Z","signature_b64":"mDTm9/TJqSeRpHUE0J9Hda9I2DGIK8DLqu6plj8oRAblH2OZUtZw3ZdoGZjVL9n2BRWRva5+xGlfBc1H42I6DQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0932c7366d09dcceb5b4fb203287d8c90e8ce38333c72a10b341191509eda14e","last_reissued_at":"2026-07-05T11:00:42.212643Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:00:42.212643Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MERGE$^3$: Efficient Evolutionary Merging on Consumer-grade GPUs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.NE","authors_text":"Adrian Robert Minut, Andrea Santilli, Donato Crisostomi, Emanuele Rodol\\`a, Tommaso Mencattini","submitted_at":"2025-02-09T14:24:16Z","abstract_excerpt":"Evolutionary model merging enables the creation of high-performing multi-task models but remains computationally prohibitive for consumer hardware. We introduce MERGE$^3$, an efficient framework that makes evolutionary merging feasible on a single GPU by reducing fitness computation costs 50$\\times$ while preserving performance. MERGE$^3$ achieves this by Extracting a reduced dataset for evaluation, Estimating model abilities using Item Response Theory (IRT), and Evolving optimal merges via IRT-based performance estimators. Our method enables state-of-the-art multilingual and cross-lingual mer"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.10436","kind":"arxiv","version":4},"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/2502.10436/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":"2502.10436","created_at":"2026-07-05T11:00:42.212699+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.10436v4","created_at":"2026-07-05T11:00:42.212699+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.10436","created_at":"2026-07-05T11:00:42.212699+00:00"},{"alias_kind":"pith_short_12","alias_value":"BEZMONTNBHOM","created_at":"2026-07-05T11:00:42.212699+00:00"},{"alias_kind":"pith_short_16","alias_value":"BEZMONTNBHOM5NNU","created_at":"2026-07-05T11:00:42.212699+00:00"},{"alias_kind":"pith_short_8","alias_value":"BEZMONTN","created_at":"2026-07-05T11:00:42.212699+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.01580","citing_title":"Model Merging: Foundations and Algorithms","ref_index":116,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE","json":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE.json","graph_json":"https://pith.science/api/pith-number/BEZMONTNBHOM5NNU7MQDFB6YZE/graph.json","events_json":"https://pith.science/api/pith-number/BEZMONTNBHOM5NNU7MQDFB6YZE/events.json","paper":"https://pith.science/paper/BEZMONTN"},"agent_actions":{"view_html":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE","download_json":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE.json","view_paper":"https://pith.science/paper/BEZMONTN","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.10436&json=true","fetch_graph":"https://pith.science/api/pith-number/BEZMONTNBHOM5NNU7MQDFB6YZE/graph.json","fetch_events":"https://pith.science/api/pith-number/BEZMONTNBHOM5NNU7MQDFB6YZE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE/action/storage_attestation","attest_author":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE/action/author_attestation","sign_citation":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE/action/citation_signature","submit_replication":"https://pith.science/pith/BEZMONTNBHOM5NNU7MQDFB6YZE/action/replication_record"}},"created_at":"2026-07-05T11:00:42.212699+00:00","updated_at":"2026-07-05T11:00:42.212699+00:00"}