{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:KJK65QQKIH55ZEUFFEDM57J7RI","short_pith_number":"pith:KJK65QQK","schema_version":"1.0","canonical_sha256":"5255eec20a41fbdc92852906cefd3f8a2474a63b6c1873373b49fa59144cca55","source":{"kind":"arxiv","id":"2406.11753","version":3},"attestation_state":"computed","paper":{"title":"A Semantic-Aware Layer-Freezing Approach to Computation-Efficient Fine-Tuning of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aldeida Aleti, Chunyang Chen, Hongyu Zhang, Jian Gu","submitted_at":"2024-06-17T17:13:08Z","abstract_excerpt":"Finetuning language models (LMs) is crucial for adapting the models to downstream data and tasks. However, full finetuning is usually costly. Existing work, such as parameter-efficient finetuning (PEFT), often focuses on \\textit{how to finetune} but neglects the issue of \\textit{where to finetune}. As a pioneering work on reducing the cost of backpropagation (at the layer level) by answering where to finetune, we conduct a semantic analysis of the LM inference process. We first propose using transition traces of the latent representation to compute deviations (or loss). Then, using a derived f"},"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":"2406.11753","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-06-17T17:13:08Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"b53e69677dd2f001872252684ee15e15817327f405fe48d5d16c0bd6e3855296","abstract_canon_sha256":"368bff7f68d2dff435b93b21020696655d6f1e51b7b5aef2265ff142ee679b25"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:13:10.709332Z","signature_b64":"BUmMS6oQrK8IlpXENzdS3JDcZM96HsL3EdBX3MCXoFoYF7Xz6WBe5gjBXv/E5ACMQwc7VTDvt0OWKimPPhyLBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5255eec20a41fbdc92852906cefd3f8a2474a63b6c1873373b49fa59144cca55","last_reissued_at":"2026-07-05T11:13:10.708812Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:13:10.708812Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"A Semantic-Aware Layer-Freezing Approach to Computation-Efficient Fine-Tuning of Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Aldeida Aleti, Chunyang Chen, Hongyu Zhang, Jian Gu","submitted_at":"2024-06-17T17:13:08Z","abstract_excerpt":"Finetuning language models (LMs) is crucial for adapting the models to downstream data and tasks. However, full finetuning is usually costly. Existing work, such as parameter-efficient finetuning (PEFT), often focuses on \\textit{how to finetune} but neglects the issue of \\textit{where to finetune}. As a pioneering work on reducing the cost of backpropagation (at the layer level) by answering where to finetune, we conduct a semantic analysis of the LM inference process. We first propose using transition traces of the latent representation to compute deviations (or loss). Then, using a derived f"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.11753","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/2406.11753/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":"2406.11753","created_at":"2026-07-05T11:13:10.708876+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.11753v3","created_at":"2026-07-05T11:13:10.708876+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.11753","created_at":"2026-07-05T11:13:10.708876+00:00"},{"alias_kind":"pith_short_12","alias_value":"KJK65QQKIH55","created_at":"2026-07-05T11:13:10.708876+00:00"},{"alias_kind":"pith_short_16","alias_value":"KJK65QQKIH55ZEUF","created_at":"2026-07-05T11:13:10.708876+00:00"},{"alias_kind":"pith_short_8","alias_value":"KJK65QQK","created_at":"2026-07-05T11:13:10.708876+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2506.23639","citing_title":"Unified Multimodal Understanding via Byte-Pair Visual Encoding","ref_index":48,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI","json":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI.json","graph_json":"https://pith.science/api/pith-number/KJK65QQKIH55ZEUFFEDM57J7RI/graph.json","events_json":"https://pith.science/api/pith-number/KJK65QQKIH55ZEUFFEDM57J7RI/events.json","paper":"https://pith.science/paper/KJK65QQK"},"agent_actions":{"view_html":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI","download_json":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI.json","view_paper":"https://pith.science/paper/KJK65QQK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.11753&json=true","fetch_graph":"https://pith.science/api/pith-number/KJK65QQKIH55ZEUFFEDM57J7RI/graph.json","fetch_events":"https://pith.science/api/pith-number/KJK65QQKIH55ZEUFFEDM57J7RI/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI/action/storage_attestation","attest_author":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI/action/author_attestation","sign_citation":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI/action/citation_signature","submit_replication":"https://pith.science/pith/KJK65QQKIH55ZEUFFEDM57J7RI/action/replication_record"}},"created_at":"2026-07-05T11:13:10.708876+00:00","updated_at":"2026-07-05T11:13:10.708876+00:00"}