{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LJRWFQDZT7HTUEIHAGCEG7XBJN","short_pith_number":"pith:LJRWFQDZ","schema_version":"1.0","canonical_sha256":"5a6362c0799fcf3a11070184437ee14b5fe26a1956fa66478a33b6d1f955a526","source":{"kind":"arxiv","id":"2310.00867","version":3},"attestation_state":"computed","paper":{"title":"Do Compressed LLMs Forget Knowledge? An Experimental Study with Practical Implications","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Duc N.M Hoang, Minsik Cho, Mohammad Rastegari, Thomas Merth, Zhangyang Wang","submitted_at":"2023-10-02T03:12:06Z","abstract_excerpt":"Compressing Large Language Models (LLMs) often leads to reduced performance, especially for knowledge-intensive tasks. In this work, we dive into how compression damages LLMs' inherent knowledge and the possible remedies. We start by proposing two conjectures on the nature of the damage: one is certain knowledge being forgotten (or erased) after LLM compression, hence necessitating the compressed model to (re)learn from data with additional parameters; the other presumes that knowledge is internally displaced and hence one requires merely \"inference re-direction\" with input-side augmentation s"},"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":"2310.00867","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2023-10-02T03:12:06Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"f5f004cc530818269e1ee83885225c0bfb1f386b0abff2d40c4efc4515e107af","abstract_canon_sha256":"e37ad1db8b5ea754fd516c53afeefa1029cb9b8c81c4d51451247a7cc436ad4f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:45:49.127724Z","signature_b64":"9yBSaa0klYcefCBXlZ2Lq7mQscy69RYKypuGXp6rjg0m3d76osipA2i+NBk/LnlOd7N1H2Ggp2t9/Hwy4ncuCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5a6362c0799fcf3a11070184437ee14b5fe26a1956fa66478a33b6d1f955a526","last_reissued_at":"2026-07-05T07:45:49.127242Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:45:49.127242Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do Compressed LLMs Forget Knowledge? An Experimental Study with Practical Implications","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Duc N.M Hoang, Minsik Cho, Mohammad Rastegari, Thomas Merth, Zhangyang Wang","submitted_at":"2023-10-02T03:12:06Z","abstract_excerpt":"Compressing Large Language Models (LLMs) often leads to reduced performance, especially for knowledge-intensive tasks. In this work, we dive into how compression damages LLMs' inherent knowledge and the possible remedies. We start by proposing two conjectures on the nature of the damage: one is certain knowledge being forgotten (or erased) after LLM compression, hence necessitating the compressed model to (re)learn from data with additional parameters; the other presumes that knowledge is internally displaced and hence one requires merely \"inference re-direction\" with input-side augmentation s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2310.00867","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/2310.00867/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":"2310.00867","created_at":"2026-07-05T07:45:49.127301+00:00"},{"alias_kind":"arxiv_version","alias_value":"2310.00867v3","created_at":"2026-07-05T07:45:49.127301+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2310.00867","created_at":"2026-07-05T07:45:49.127301+00:00"},{"alias_kind":"pith_short_12","alias_value":"LJRWFQDZT7HT","created_at":"2026-07-05T07:45:49.127301+00:00"},{"alias_kind":"pith_short_16","alias_value":"LJRWFQDZT7HTUEIH","created_at":"2026-07-05T07:45:49.127301+00:00"},{"alias_kind":"pith_short_8","alias_value":"LJRWFQDZ","created_at":"2026-07-05T07:45:49.127301+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03328","citing_title":"Averaged Evaluation Masks Capability Trade-Offs: Multi-Source Calibration for High-Sparsity LLM Pruning","ref_index":19,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN","json":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN.json","graph_json":"https://pith.science/api/pith-number/LJRWFQDZT7HTUEIHAGCEG7XBJN/graph.json","events_json":"https://pith.science/api/pith-number/LJRWFQDZT7HTUEIHAGCEG7XBJN/events.json","paper":"https://pith.science/paper/LJRWFQDZ"},"agent_actions":{"view_html":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN","download_json":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN.json","view_paper":"https://pith.science/paper/LJRWFQDZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2310.00867&json=true","fetch_graph":"https://pith.science/api/pith-number/LJRWFQDZT7HTUEIHAGCEG7XBJN/graph.json","fetch_events":"https://pith.science/api/pith-number/LJRWFQDZT7HTUEIHAGCEG7XBJN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN/action/storage_attestation","attest_author":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN/action/author_attestation","sign_citation":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN/action/citation_signature","submit_replication":"https://pith.science/pith/LJRWFQDZT7HTUEIHAGCEG7XBJN/action/replication_record"}},"created_at":"2026-07-05T07:45:49.127301+00:00","updated_at":"2026-07-05T07:45:49.127301+00:00"}