{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:UU3ZMT4KGAAIATHWRRPFCJZTVP","short_pith_number":"pith:UU3ZMT4K","schema_version":"1.0","canonical_sha256":"a537964f8a3000804cf68c5e512733abed7eb90f924a0b9b9ed7880acb26d1a7","source":{"kind":"arxiv","id":"2606.02823","version":1},"attestation_state":"computed","paper":{"title":"Qift: Shift-Friendly No-Zero W2 Post-Training Quantization for Rotated W2A4/KV4 LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chia-Chi Tsai, Chi-Wei Huang","submitted_at":"2026-06-01T19:40:32Z","abstract_excerpt":"Two-bit weight quantization is attractive for memory-efficient LLM inference, but the standard W2 level set {-2,-1,0,+1} often collapses under aggressive W2A4/KV4 settings. We study the scalar level-set geometry of two-bit weights in a Hadamard-rotated quantization pipeline. Conventional asymmetric W2 substantially improves over the standard level set, indicating that W2A4 failure is not only a bit-width problem but also a reconstruction-level problem. Across all 224 linear modules in each of LLaMA-2-7B and LLaMA-3.1-8B, pretrained weights are already nearly zero-centered, while Hadamard rotat"},"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":"2606.02823","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-01T19:40:32Z","cross_cats_sorted":[],"title_canon_sha256":"219e3fae004dc4f894c8724b6909ab18a38b719f3ae09be8110ffa1d49084305","abstract_canon_sha256":"41ea3f39dea6e541dad54836e889f7fa90bbdd216ac6dac25d7272b78de5bf08"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-03T01:05:23.911543Z","signature_b64":"MIncAwzKyOGtkj7I3eDjsnc2ujpGnWIG7zFmYYcd1Rj3WY3S9GLySZ1lz6IYu/obOfGLWF0s8FWC+yCoLc9LDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a537964f8a3000804cf68c5e512733abed7eb90f924a0b9b9ed7880acb26d1a7","last_reissued_at":"2026-06-03T01:05:23.911167Z","signature_status":"signed_v1","first_computed_at":"2026-06-03T01:05:23.911167Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Qift: Shift-Friendly No-Zero W2 Post-Training Quantization for Rotated W2A4/KV4 LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Chia-Chi Tsai, Chi-Wei Huang","submitted_at":"2026-06-01T19:40:32Z","abstract_excerpt":"Two-bit weight quantization is attractive for memory-efficient LLM inference, but the standard W2 level set {-2,-1,0,+1} often collapses under aggressive W2A4/KV4 settings. We study the scalar level-set geometry of two-bit weights in a Hadamard-rotated quantization pipeline. Conventional asymmetric W2 substantially improves over the standard level set, indicating that W2A4 failure is not only a bit-width problem but also a reconstruction-level problem. Across all 224 linear modules in each of LLaMA-2-7B and LLaMA-3.1-8B, pretrained weights are already nearly zero-centered, while Hadamard rotat"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.02823","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/2606.02823/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":"2606.02823","created_at":"2026-06-03T01:05:23.911230+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.02823v1","created_at":"2026-06-03T01:05:23.911230+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.02823","created_at":"2026-06-03T01:05:23.911230+00:00"},{"alias_kind":"pith_short_12","alias_value":"UU3ZMT4KGAAI","created_at":"2026-06-03T01:05:23.911230+00:00"},{"alias_kind":"pith_short_16","alias_value":"UU3ZMT4KGAAIATHW","created_at":"2026-06-03T01:05:23.911230+00:00"},{"alias_kind":"pith_short_8","alias_value":"UU3ZMT4K","created_at":"2026-06-03T01:05:23.911230+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP","json":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP.json","graph_json":"https://pith.science/api/pith-number/UU3ZMT4KGAAIATHWRRPFCJZTVP/graph.json","events_json":"https://pith.science/api/pith-number/UU3ZMT4KGAAIATHWRRPFCJZTVP/events.json","paper":"https://pith.science/paper/UU3ZMT4K"},"agent_actions":{"view_html":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP","download_json":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP.json","view_paper":"https://pith.science/paper/UU3ZMT4K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.02823&json=true","fetch_graph":"https://pith.science/api/pith-number/UU3ZMT4KGAAIATHWRRPFCJZTVP/graph.json","fetch_events":"https://pith.science/api/pith-number/UU3ZMT4KGAAIATHWRRPFCJZTVP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP/action/storage_attestation","attest_author":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP/action/author_attestation","sign_citation":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP/action/citation_signature","submit_replication":"https://pith.science/pith/UU3ZMT4KGAAIATHWRRPFCJZTVP/action/replication_record"}},"created_at":"2026-06-03T01:05:23.911230+00:00","updated_at":"2026-06-03T01:05:23.911230+00:00"}