{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:GBSURCVPE67IUCRE62F3RFA5HP","short_pith_number":"pith:GBSURCVP","schema_version":"1.0","canonical_sha256":"3065488aaf27be8a0a24f68bb8941d3bd2b0ee60b9a10b162b7285ecc6787bb0","source":{"kind":"arxiv","id":"2506.09104","version":1},"attestation_state":"computed","paper":{"title":"Unifying Block-wise PTQ and Distillation-based QAT for Progressive Quantization toward 2-bit Instruction-Tuned LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"An Chen, Jaeseong You, Jung Hyun Lee, Seungjae Shin, Vinnam Kim","submitted_at":"2025-06-10T16:26:32Z","abstract_excerpt":"As the rapid scaling of large language models (LLMs) poses significant challenges for deployment on resource-constrained devices, there is growing interest in extremely low-bit quantization, such as 2-bit. Although prior works have shown that 2-bit large models are pareto-optimal over their 4-bit smaller counterparts in both accuracy and latency, these advancements have been limited to pre-trained LLMs and have not yet been extended to instruction-tuned models. To bridge this gap, we propose Unified Progressive Quantization (UPQ)$-$a novel progressive quantization framework (FP16$\\rightarrow$I"},"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":"2506.09104","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-10T16:26:32Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"d6e621ec92b51f7018d76acbc3af4a70384e2c2a736dcb6d71cfd3a1336b4ff8","abstract_canon_sha256":"a06f280c4525dec3cc211de5220bad38e59dae87d25b7948e16820a88b1b8cbe"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:19:31.566137Z","signature_b64":"KNIX5Gs99BMNOFYrboMfbHSbFVg7RXmDPb/wJlaJ2OMM/2s3Xpo/RQMOv0M2bYz6iAl65ttr5kGpfnumgkS+Cw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3065488aaf27be8a0a24f68bb8941d3bd2b0ee60b9a10b162b7285ecc6787bb0","last_reissued_at":"2026-07-05T11:19:31.565674Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:19:31.565674Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Unifying Block-wise PTQ and Distillation-based QAT for Progressive Quantization toward 2-bit Instruction-Tuned LLMs","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"An Chen, Jaeseong You, Jung Hyun Lee, Seungjae Shin, Vinnam Kim","submitted_at":"2025-06-10T16:26:32Z","abstract_excerpt":"As the rapid scaling of large language models (LLMs) poses significant challenges for deployment on resource-constrained devices, there is growing interest in extremely low-bit quantization, such as 2-bit. Although prior works have shown that 2-bit large models are pareto-optimal over their 4-bit smaller counterparts in both accuracy and latency, these advancements have been limited to pre-trained LLMs and have not yet been extended to instruction-tuned models. To bridge this gap, we propose Unified Progressive Quantization (UPQ)$-$a novel progressive quantization framework (FP16$\\rightarrow$I"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.09104","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/2506.09104/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":"2506.09104","created_at":"2026-07-05T11:19:31.565732+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.09104v1","created_at":"2026-07-05T11:19:31.565732+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.09104","created_at":"2026-07-05T11:19:31.565732+00:00"},{"alias_kind":"pith_short_12","alias_value":"GBSURCVPE67I","created_at":"2026-07-05T11:19:31.565732+00:00"},{"alias_kind":"pith_short_16","alias_value":"GBSURCVPE67IUCRE","created_at":"2026-07-05T11:19:31.565732+00:00"},{"alias_kind":"pith_short_8","alias_value":"GBSURCVP","created_at":"2026-07-05T11:19:31.565732+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.29128","citing_title":"Apertus LLM Family Expansion via Distillation and Quantization","ref_index":8,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP","json":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP.json","graph_json":"https://pith.science/api/pith-number/GBSURCVPE67IUCRE62F3RFA5HP/graph.json","events_json":"https://pith.science/api/pith-number/GBSURCVPE67IUCRE62F3RFA5HP/events.json","paper":"https://pith.science/paper/GBSURCVP"},"agent_actions":{"view_html":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP","download_json":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP.json","view_paper":"https://pith.science/paper/GBSURCVP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.09104&json=true","fetch_graph":"https://pith.science/api/pith-number/GBSURCVPE67IUCRE62F3RFA5HP/graph.json","fetch_events":"https://pith.science/api/pith-number/GBSURCVPE67IUCRE62F3RFA5HP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP/action/storage_attestation","attest_author":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP/action/author_attestation","sign_citation":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP/action/citation_signature","submit_replication":"https://pith.science/pith/GBSURCVPE67IUCRE62F3RFA5HP/action/replication_record"}},"created_at":"2026-07-05T11:19:31.565732+00:00","updated_at":"2026-07-05T11:19:31.565732+00:00"}