{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:LQ2NUAV6LSI7QJZP2ASNIKJTAW","short_pith_number":"pith:LQ2NUAV6","schema_version":"1.0","canonical_sha256":"5c34da02be5c91f8272fd024d42933059533d2f75180c97858dc1ed65d03eb3a","source":{"kind":"arxiv","id":"2405.18137","version":2},"attestation_state":"computed","paper":{"title":"Exploiting LLM Quantization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Jingxuan He, Kazuki Egashira, Mark Vero, Martin Vechev, Robin Staab","submitted_at":"2024-05-28T12:51:01Z","abstract_excerpt":"Quantization leverages lower-precision weights to reduce the memory usage of large language models (LLMs) and is a key technique for enabling their deployment on commodity hardware. While LLM quantization's impact on utility has been extensively explored, this work for the first time studies its adverse effects from a security perspective. We reveal that widely used quantization methods can be exploited to produce a harmful quantized LLM, even though the full-precision counterpart appears benign, potentially tricking users into deploying the malicious quantized model. We demonstrate this threa"},"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":"2405.18137","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-28T12:51:01Z","cross_cats_sorted":["cs.AI","cs.CR"],"title_canon_sha256":"5c06bd467bfbc185df8755b0342891f02b0ddc1adcc66c2558aa42104504f910","abstract_canon_sha256":"6130f12b75eef4795edd9b43ec43a92577952330845e095131459e60394c1f18"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:30:15.334083Z","signature_b64":"fwtCrOvFwrq/0izu0keexUW1yWz3ZSEQBWVkoX/7LRo9NZKBy3/rTO0RsWRk74+BdvS72UHJ1Xsn9kZmGKDTBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5c34da02be5c91f8272fd024d42933059533d2f75180c97858dc1ed65d03eb3a","last_reissued_at":"2026-07-05T09:30:15.333534Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:30:15.333534Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Exploiting LLM Quantization","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CR"],"primary_cat":"cs.LG","authors_text":"Jingxuan He, Kazuki Egashira, Mark Vero, Martin Vechev, Robin Staab","submitted_at":"2024-05-28T12:51:01Z","abstract_excerpt":"Quantization leverages lower-precision weights to reduce the memory usage of large language models (LLMs) and is a key technique for enabling their deployment on commodity hardware. While LLM quantization's impact on utility has been extensively explored, this work for the first time studies its adverse effects from a security perspective. We reveal that widely used quantization methods can be exploited to produce a harmful quantized LLM, even though the full-precision counterpart appears benign, potentially tricking users into deploying the malicious quantized model. We demonstrate this threa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.18137","kind":"arxiv","version":2},"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/2405.18137/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":"2405.18137","created_at":"2026-07-05T09:30:15.333604+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.18137v2","created_at":"2026-07-05T09:30:15.333604+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.18137","created_at":"2026-07-05T09:30:15.333604+00:00"},{"alias_kind":"pith_short_12","alias_value":"LQ2NUAV6LSI7","created_at":"2026-07-05T09:30:15.333604+00:00"},{"alias_kind":"pith_short_16","alias_value":"LQ2NUAV6LSI7QJZP","created_at":"2026-07-05T09:30:15.333604+00:00"},{"alias_kind":"pith_short_8","alias_value":"LQ2NUAV6","created_at":"2026-07-05T09:30:15.333604+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.10154","citing_title":"Quality Is Not a Safety Proxy Under Quantization","ref_index":13,"is_internal_anchor":false},{"citing_arxiv_id":"2605.27763","citing_title":"A Paired Testing Protocol for Batch-Conditioned Refusal Robustness in LLM Serving","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2411.10656","citing_title":"Precision or Peril: A PoC of Python Code Quality from Quantized Large Language Models","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2505.12601","citing_title":"Rethinking Predictive Modeling for LLM Routing: When Simple kNN Beats Complex Learned Routers","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW","json":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW.json","graph_json":"https://pith.science/api/pith-number/LQ2NUAV6LSI7QJZP2ASNIKJTAW/graph.json","events_json":"https://pith.science/api/pith-number/LQ2NUAV6LSI7QJZP2ASNIKJTAW/events.json","paper":"https://pith.science/paper/LQ2NUAV6"},"agent_actions":{"view_html":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW","download_json":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW.json","view_paper":"https://pith.science/paper/LQ2NUAV6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.18137&json=true","fetch_graph":"https://pith.science/api/pith-number/LQ2NUAV6LSI7QJZP2ASNIKJTAW/graph.json","fetch_events":"https://pith.science/api/pith-number/LQ2NUAV6LSI7QJZP2ASNIKJTAW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW/action/storage_attestation","attest_author":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW/action/author_attestation","sign_citation":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW/action/citation_signature","submit_replication":"https://pith.science/pith/LQ2NUAV6LSI7QJZP2ASNIKJTAW/action/replication_record"}},"created_at":"2026-07-05T09:30:15.333604+00:00","updated_at":"2026-07-05T09:30:15.333604+00:00"}