{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:NWHMLINBOHOWXGXKN5LC2TWWAP","short_pith_number":"pith:NWHMLINB","schema_version":"1.0","canonical_sha256":"6d8ec5a1a171dd6b9aea6f562d4ed603ddf290a03cc584243d98beb9f9520487","source":{"kind":"arxiv","id":"2406.08903","version":3},"attestation_state":"computed","paper":{"title":"Delta-CoMe: Training-Free Delta-Compression with Mixed-Precision for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baobao Chang, Bowen Ping, Hanqing Wang, Maosong Sun, Shuo Wang, Xu Han, Yukun Yan, Yun Chen, Yuzhuang Xu, Zhiyuan Liu","submitted_at":"2024-06-13T07:57:27Z","abstract_excerpt":"Fine-tuning is a crucial process for adapting large language models (LLMs) to diverse applications. In certain scenarios, such as multi-tenant serving, deploying multiple LLMs becomes necessary to meet complex demands. Recent studies suggest decomposing a fine-tuned LLM into a base model and corresponding delta weights, which are then compressed using low-rank or low-bit approaches to reduce costs. In this work, we observe that existing low-rank and low-bit compression methods can significantly harm the model performance for task-specific fine-tuned LLMs (e.g., WizardMath for math problems). M"},"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":"2406.08903","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-13T07:57:27Z","cross_cats_sorted":[],"title_canon_sha256":"cc45bf4fcc85b598a99b452d992cc3e4de07996f91db5fa868de336f99f1c97f","abstract_canon_sha256":"ea9c4756269ce4d62319a26aa0d606a86b5ef0f1a38a271d3cb3d0b667801403"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:40:18.116772Z","signature_b64":"qfwyHpYlpco1G4H+CcxpNSBkb4QDxlDuA5+/RrJkLn1OCfrjRNflhPdS7y9ffmIEKxLiH53C0Utd4RluX43jDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6d8ec5a1a171dd6b9aea6f562d4ed603ddf290a03cc584243d98beb9f9520487","last_reissued_at":"2026-07-05T09:40:18.116211Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:40:18.116211Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Delta-CoMe: Training-Free Delta-Compression with Mixed-Precision for Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Baobao Chang, Bowen Ping, Hanqing Wang, Maosong Sun, Shuo Wang, Xu Han, Yukun Yan, Yun Chen, Yuzhuang Xu, Zhiyuan Liu","submitted_at":"2024-06-13T07:57:27Z","abstract_excerpt":"Fine-tuning is a crucial process for adapting large language models (LLMs) to diverse applications. In certain scenarios, such as multi-tenant serving, deploying multiple LLMs becomes necessary to meet complex demands. Recent studies suggest decomposing a fine-tuned LLM into a base model and corresponding delta weights, which are then compressed using low-rank or low-bit approaches to reduce costs. In this work, we observe that existing low-rank and low-bit compression methods can significantly harm the model performance for task-specific fine-tuned LLMs (e.g., WizardMath for math problems). M"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.08903","kind":"arxiv","version":3},"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/2406.08903/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":"2406.08903","created_at":"2026-07-05T09:40:18.116268+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.08903v3","created_at":"2026-07-05T09:40:18.116268+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.08903","created_at":"2026-07-05T09:40:18.116268+00:00"},{"alias_kind":"pith_short_12","alias_value":"NWHMLINBOHOW","created_at":"2026-07-05T09:40:18.116268+00:00"},{"alias_kind":"pith_short_16","alias_value":"NWHMLINBOHOWXGXK","created_at":"2026-07-05T09:40:18.116268+00:00"},{"alias_kind":"pith_short_8","alias_value":"NWHMLINB","created_at":"2026-07-05T09:40:18.116268+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2502.02095","citing_title":"LongDPO: Unlock Better Long-form Generation Abilities for LLMs via Critique-augmented Stepwise Information","ref_index":29,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP","json":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP.json","graph_json":"https://pith.science/api/pith-number/NWHMLINBOHOWXGXKN5LC2TWWAP/graph.json","events_json":"https://pith.science/api/pith-number/NWHMLINBOHOWXGXKN5LC2TWWAP/events.json","paper":"https://pith.science/paper/NWHMLINB"},"agent_actions":{"view_html":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP","download_json":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP.json","view_paper":"https://pith.science/paper/NWHMLINB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.08903&json=true","fetch_graph":"https://pith.science/api/pith-number/NWHMLINBOHOWXGXKN5LC2TWWAP/graph.json","fetch_events":"https://pith.science/api/pith-number/NWHMLINBOHOWXGXKN5LC2TWWAP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP/action/storage_attestation","attest_author":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP/action/author_attestation","sign_citation":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP/action/citation_signature","submit_replication":"https://pith.science/pith/NWHMLINBOHOWXGXKN5LC2TWWAP/action/replication_record"}},"created_at":"2026-07-05T09:40:18.116268+00:00","updated_at":"2026-07-05T09:40:18.116268+00:00"}