{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FI6T3EVVCFAGEYI22EGRW5CA7M","short_pith_number":"pith:FI6T3EVV","schema_version":"1.0","canonical_sha256":"2a3d3d92b5114062611ad10d1b7440fb3b98dc9f7bdcd4986e0899ebca28c689","source":{"kind":"arxiv","id":"2605.18822","version":1},"attestation_state":"computed","paper":{"title":"Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengqian Zhang, Kyumin Lee, Wei Zhu","submitted_at":"2026-05-12T15:11:44Z","abstract_excerpt":"Post-training has become essential for adapting large language models (LLMs) to complex downstream behaviors, including instruction following, preference alignment, and multi-step reasoning. Reinforcement learning with verifiable rewards (RLVR) has recently emerged as a particularly effective post-training paradigm for improving reasoning capabilities, with critic-free algorithms such as GRPO and GSPO enabling scalable optimization. However, RLVR post-training with full fine-tuning (FFT) requires substantial GPU memory and incurs high training costs. Although parameter-efficient fine-tuning (P"},"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":"2605.18822","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-12T15:11:44Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a6c4091268bcecd5710c8ba12bdd1529ef18d968a4ced318f54ec9735b178449","abstract_canon_sha256":"4661cc63dcaae31c520668568a427e35d631b249b5b9a671b495da10976855e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:06:24.402861Z","signature_b64":"y7M7zu6XasQVKfKrmVbdchwa4lv35FUikhTbGKp4fByhhqCasIGOcsccHUAvIQ9b4k5L5IvffGr1m2S72XivAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2a3d3d92b5114062611ad10d1b7440fb3b98dc9f7bdcd4986e0899ebca28c689","last_reissued_at":"2026-05-20T00:06:24.402060Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:06:24.402060Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Hybrid-LoRA: Bridging Full Fine-Tuning and Low-Rank Adaptation for Post-Training","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Chengqian Zhang, Kyumin Lee, Wei Zhu","submitted_at":"2026-05-12T15:11:44Z","abstract_excerpt":"Post-training has become essential for adapting large language models (LLMs) to complex downstream behaviors, including instruction following, preference alignment, and multi-step reasoning. Reinforcement learning with verifiable rewards (RLVR) has recently emerged as a particularly effective post-training paradigm for improving reasoning capabilities, with critic-free algorithms such as GRPO and GSPO enabling scalable optimization. However, RLVR post-training with full fine-tuning (FFT) requires substantial GPU memory and incurs high training costs. Although parameter-efficient fine-tuning (P"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.18822","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/2605.18822/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":"2605.18822","created_at":"2026-05-20T00:06:24.402199+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.18822v1","created_at":"2026-05-20T00:06:24.402199+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.18822","created_at":"2026-05-20T00:06:24.402199+00:00"},{"alias_kind":"pith_short_12","alias_value":"FI6T3EVVCFAG","created_at":"2026-05-20T00:06:24.402199+00:00"},{"alias_kind":"pith_short_16","alias_value":"FI6T3EVVCFAGEYI2","created_at":"2026-05-20T00:06:24.402199+00:00"},{"alias_kind":"pith_short_8","alias_value":"FI6T3EVV","created_at":"2026-05-20T00:06:24.402199+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/FI6T3EVVCFAGEYI22EGRW5CA7M","json":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M.json","graph_json":"https://pith.science/api/pith-number/FI6T3EVVCFAGEYI22EGRW5CA7M/graph.json","events_json":"https://pith.science/api/pith-number/FI6T3EVVCFAGEYI22EGRW5CA7M/events.json","paper":"https://pith.science/paper/FI6T3EVV"},"agent_actions":{"view_html":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M","download_json":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M.json","view_paper":"https://pith.science/paper/FI6T3EVV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.18822&json=true","fetch_graph":"https://pith.science/api/pith-number/FI6T3EVVCFAGEYI22EGRW5CA7M/graph.json","fetch_events":"https://pith.science/api/pith-number/FI6T3EVVCFAGEYI22EGRW5CA7M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M/action/storage_attestation","attest_author":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M/action/author_attestation","sign_citation":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M/action/citation_signature","submit_replication":"https://pith.science/pith/FI6T3EVVCFAGEYI22EGRW5CA7M/action/replication_record"}},"created_at":"2026-05-20T00:06:24.402199+00:00","updated_at":"2026-05-20T00:06:24.402199+00:00"}