{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:BU4FEKSQ4J43HKIXYSFRDAB7OV","short_pith_number":"pith:BU4FEKSQ","schema_version":"1.0","canonical_sha256":"0d38522a50e279b3a917c48b11803f75614bec7657fc54ee4685df9037ccd7ca","source":{"kind":"arxiv","id":"2405.14852","version":2},"attestation_state":"computed","paper":{"title":"PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Denis Kuznedelev, Denis Mazur, Ivan Ilin, Kai Yi, Konstantin Burlachenko, Peter Richtarik, Vladimir Malinovskii","submitted_at":"2024-05-23T17:57:04Z","abstract_excerpt":"There has been significant interest in \"extreme\" compression of large language models (LLMs), i.e., to 1-2 bits per parameter, which allows such models to be executed efficiently on resource-constrained devices. Existing work focused on improved one-shot quantization techniques and weight representations; yet, purely post-training approaches are reaching diminishing returns in terms of the accuracy-vs-bit-width trade-off. State-of-the-art quantization methods such as QuIP# and AQLM include fine-tuning (part of) the compressed parameters over a limited amount of calibration data; however, such "},"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":"2405.14852","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-05-23T17:57:04Z","cross_cats_sorted":[],"title_canon_sha256":"01da8e2ad74c9047f681bda67e3228db1a9ed7f5237020fd6cf6ad052c37e21e","abstract_canon_sha256":"b42ce9918d65447cea1302cb0158abac5848d921dd7f820a37788f0123454a38"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:25:04.555007Z","signature_b64":"9H9tPVb/Qj5MpmNig4EXXKFUqZ5T9uNjEvEiM2Is6pehFyHpw2S5+r1MB1/+A5mQRfCX3GxviXFzbuq1+qynCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0d38522a50e279b3a917c48b11803f75614bec7657fc54ee4685df9037ccd7ca","last_reissued_at":"2026-07-05T08:25:04.554462Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:25:04.554462Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"PV-Tuning: Beyond Straight-Through Estimation for Extreme LLM Compression","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Dan Alistarh, Denis Kuznedelev, Denis Mazur, Ivan Ilin, Kai Yi, Konstantin Burlachenko, Peter Richtarik, Vladimir Malinovskii","submitted_at":"2024-05-23T17:57:04Z","abstract_excerpt":"There has been significant interest in \"extreme\" compression of large language models (LLMs), i.e., to 1-2 bits per parameter, which allows such models to be executed efficiently on resource-constrained devices. Existing work focused on improved one-shot quantization techniques and weight representations; yet, purely post-training approaches are reaching diminishing returns in terms of the accuracy-vs-bit-width trade-off. State-of-the-art quantization methods such as QuIP# and AQLM include fine-tuning (part of) the compressed parameters over a limited amount of calibration data; however, such "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2405.14852","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/2405.14852/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":"2405.14852","created_at":"2026-07-05T08:25:04.554532+00:00"},{"alias_kind":"arxiv_version","alias_value":"2405.14852v2","created_at":"2026-07-05T08:25:04.554532+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2405.14852","created_at":"2026-07-05T08:25:04.554532+00:00"},{"alias_kind":"pith_short_12","alias_value":"BU4FEKSQ4J43","created_at":"2026-07-05T08:25:04.554532+00:00"},{"alias_kind":"pith_short_16","alias_value":"BU4FEKSQ4J43HKIX","created_at":"2026-07-05T08:25:04.554532+00:00"},{"alias_kind":"pith_short_8","alias_value":"BU4FEKSQ","created_at":"2026-07-05T08:25:04.554532+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.09425","citing_title":"AtteConDA: Attention-Based Conflict Suppression in Multi-Condition Diffusion Models and Synthetic Data Augmentation","ref_index":51,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10655","citing_title":"BCJR-QAT: A Differentiable Relaxation of Trellis-Coded Weight Quantization","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV","json":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV.json","graph_json":"https://pith.science/api/pith-number/BU4FEKSQ4J43HKIXYSFRDAB7OV/graph.json","events_json":"https://pith.science/api/pith-number/BU4FEKSQ4J43HKIXYSFRDAB7OV/events.json","paper":"https://pith.science/paper/BU4FEKSQ"},"agent_actions":{"view_html":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV","download_json":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV.json","view_paper":"https://pith.science/paper/BU4FEKSQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2405.14852&json=true","fetch_graph":"https://pith.science/api/pith-number/BU4FEKSQ4J43HKIXYSFRDAB7OV/graph.json","fetch_events":"https://pith.science/api/pith-number/BU4FEKSQ4J43HKIXYSFRDAB7OV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV/action/storage_attestation","attest_author":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV/action/author_attestation","sign_citation":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV/action/citation_signature","submit_replication":"https://pith.science/pith/BU4FEKSQ4J43HKIXYSFRDAB7OV/action/replication_record"}},"created_at":"2026-07-05T08:25:04.554532+00:00","updated_at":"2026-07-05T08:25:04.554532+00:00"}