{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3RCX7EXDOSD7HONB3G5NZ74FTR","short_pith_number":"pith:3RCX7EXD","schema_version":"1.0","canonical_sha256":"dc457f92e37487f3b9a1d9badcff859c432d69e0f01a4a09bb247d389a576be5","source":{"kind":"arxiv","id":"2605.14844","version":1},"attestation_state":"computed","paper":{"title":"XFP: Quality-Targeted Adaptive Codebook Quantization with Sparse Outlier Separation for LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Thomas Witt","submitted_at":"2026-05-14T13:52:31Z","abstract_excerpt":"We introduce XFP, a dynamic weight quantizer for LLM inference that inverts the conventional workflow: the operator specifies reconstruction quality floors on per-channel cosine similarity (one strict floor for attention and shared experts, one lazy floor for routed-expert MoE); XFP determines codebook size, outlier budget, and packing per layer automatically -- no Hessian, no calibration data, no manual bit-width selection. Each weight matrix is decomposed into a sparse fp16 outlier residual and a dense sub-byte index tensor into a per-group learned codebook. Two storage modes share one auto-"},"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":"2605.14844","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-14T13:52:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"3cc0eb04bc13d6e338b2abc64ad4f9fc0339c997e8f471ec7587f54d036d1d32","abstract_canon_sha256":"896c33c6ad11a852e159a75b6f5863bab6532b4625bee20d529807bb678b13c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-17T23:38:56.386525Z","signature_b64":"gGGEilM/riPq8oR9QaR3BjCGDu/XEAI3OqgtdyHhV0EvbeCKjVOGgj6bJBQRa7NSGpd5+ZYUYQMRAX+BWVwWDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dc457f92e37487f3b9a1d9badcff859c432d69e0f01a4a09bb247d389a576be5","last_reissued_at":"2026-05-17T23:38:56.385801Z","signature_status":"signed_v1","first_computed_at":"2026-05-17T23:38:56.385801Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"XFP: Quality-Targeted Adaptive Codebook Quantization with Sparse Outlier Separation for LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.LG","authors_text":"Thomas Witt","submitted_at":"2026-05-14T13:52:31Z","abstract_excerpt":"We introduce XFP, a dynamic weight quantizer for LLM inference that inverts the conventional workflow: the operator specifies reconstruction quality floors on per-channel cosine similarity (one strict floor for attention and shared experts, one lazy floor for routed-expert MoE); XFP determines codebook size, outlier budget, and packing per layer automatically -- no Hessian, no calibration data, no manual bit-width selection. Each weight matrix is decomposed into a sparse fp16 outlier residual and a dense sub-byte index tensor into a per-group learned codebook. Two storage modes share one auto-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.14844","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":""},"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":"2605.14844","created_at":"2026-05-17T23:38:56.385924+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.14844v1","created_at":"2026-05-17T23:38:56.385924+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.14844","created_at":"2026-05-17T23:38:56.385924+00:00"},{"alias_kind":"pith_short_12","alias_value":"3RCX7EXDOSD7","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"3RCX7EXDOSD7HONB","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"3RCX7EXD","created_at":"2026-05-18T12:33:37.589309+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/3RCX7EXDOSD7HONB3G5NZ74FTR","json":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR.json","graph_json":"https://pith.science/api/pith-number/3RCX7EXDOSD7HONB3G5NZ74FTR/graph.json","events_json":"https://pith.science/api/pith-number/3RCX7EXDOSD7HONB3G5NZ74FTR/events.json","paper":"https://pith.science/paper/3RCX7EXD"},"agent_actions":{"view_html":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR","download_json":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR.json","view_paper":"https://pith.science/paper/3RCX7EXD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.14844&json=true","fetch_graph":"https://pith.science/api/pith-number/3RCX7EXDOSD7HONB3G5NZ74FTR/graph.json","fetch_events":"https://pith.science/api/pith-number/3RCX7EXDOSD7HONB3G5NZ74FTR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR/action/storage_attestation","attest_author":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR/action/author_attestation","sign_citation":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR/action/citation_signature","submit_replication":"https://pith.science/pith/3RCX7EXDOSD7HONB3G5NZ74FTR/action/replication_record"}},"created_at":"2026-05-17T23:38:56.385924+00:00","updated_at":"2026-05-17T23:38:56.385924+00:00"}