{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:GR4O4KEUJL5WOQZFQ4WFX2233D","short_pith_number":"pith:GR4O4KEU","schema_version":"1.0","canonical_sha256":"3478ee28944afb674325872c5beb5bd8cbfc792fb8b4fef698ac66338345063f","source":{"kind":"arxiv","id":"2608.03867","version":1},"attestation_state":"computed","paper":{"title":"Heterogeneity-Aware Microscaling for Efficient Low-Bit LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AR","authors_text":"Ben Laurie, Dennis Sylvester, Gregory Kielian, Hongyi Wu, Junyi Luo, Mehdi Saligane, Minxing Chu, Qirui Zhang, Quan Cheng, Ruichen Qi, Tai-Hao Wen, Xinting Jiang","submitted_at":"2026-08-04T16:08:20Z","abstract_excerpt":"Microscaling (MX) is now the standard for low-bit large language model (LLM) inference. Its 4-bit form MXFP4 still loses substantial accuracy, because existing MX formats fix either the element format or the precision-recovery scheme across blocks, and thus capture only limited quantization heterogeneity. Quantization heterogeneity appears at two levels: 1) across blocks, the preferred element format and precision-recovery scheme vary; 2) across operands, weights and activations require different encoding. We introduce AdaMX (Adaptive Microscaling), a heterogeneity-aware format and accelerator"},"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":"2608.03867","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.AR","submitted_at":"2026-08-04T16:08:20Z","cross_cats_sorted":[],"title_canon_sha256":"0df129d4c67c767ff4ae363aa29f3b579b8abae0c164e3cb8daa0f157bdb5440","abstract_canon_sha256":"d0e0d9a6452439ab22308f498dce2cf35b56493beb583c380b075835406ad374"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T01:37:16.960752Z","signature_b64":"cbxKO1f9+awLZ8L+PbjRotJRX3+kfyAM1ZX96I8HRARf2geoVXK2ell+xmo0ufZLq3QEh6QfnQTFM/7hEbJNCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3478ee28944afb674325872c5beb5bd8cbfc792fb8b4fef698ac66338345063f","last_reissued_at":"2026-08-05T01:37:16.958894Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T01:37:16.958894Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Heterogeneity-Aware Microscaling for Efficient Low-Bit LLM Inference","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.AR","authors_text":"Ben Laurie, Dennis Sylvester, Gregory Kielian, Hongyi Wu, Junyi Luo, Mehdi Saligane, Minxing Chu, Qirui Zhang, Quan Cheng, Ruichen Qi, Tai-Hao Wen, Xinting Jiang","submitted_at":"2026-08-04T16:08:20Z","abstract_excerpt":"Microscaling (MX) is now the standard for low-bit large language model (LLM) inference. Its 4-bit form MXFP4 still loses substantial accuracy, because existing MX formats fix either the element format or the precision-recovery scheme across blocks, and thus capture only limited quantization heterogeneity. Quantization heterogeneity appears at two levels: 1) across blocks, the preferred element format and precision-recovery scheme vary; 2) across operands, weights and activations require different encoding. We introduce AdaMX (Adaptive Microscaling), a heterogeneity-aware format and accelerator"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.03867","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/2608.03867/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":"2608.03867","created_at":"2026-08-05T01:37:16.959536+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.03867v1","created_at":"2026-08-05T01:37:16.959536+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.03867","created_at":"2026-08-05T01:37:16.959536+00:00"},{"alias_kind":"pith_short_12","alias_value":"GR4O4KEUJL5W","created_at":"2026-08-05T01:37:16.959536+00:00"},{"alias_kind":"pith_short_16","alias_value":"GR4O4KEUJL5WOQZF","created_at":"2026-08-05T01:37:16.959536+00:00"},{"alias_kind":"pith_short_8","alias_value":"GR4O4KEU","created_at":"2026-08-05T01:37:16.959536+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/GR4O4KEUJL5WOQZFQ4WFX2233D","json":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D.json","graph_json":"https://pith.science/api/pith-number/GR4O4KEUJL5WOQZFQ4WFX2233D/graph.json","events_json":"https://pith.science/api/pith-number/GR4O4KEUJL5WOQZFQ4WFX2233D/events.json","paper":"https://pith.science/paper/GR4O4KEU"},"agent_actions":{"view_html":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D","download_json":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D.json","view_paper":"https://pith.science/paper/GR4O4KEU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.03867&json=true","fetch_graph":"https://pith.science/api/pith-number/GR4O4KEUJL5WOQZFQ4WFX2233D/graph.json","fetch_events":"https://pith.science/api/pith-number/GR4O4KEUJL5WOQZFQ4WFX2233D/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D/action/storage_attestation","attest_author":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D/action/author_attestation","sign_citation":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D/action/citation_signature","submit_replication":"https://pith.science/pith/GR4O4KEUJL5WOQZFQ4WFX2233D/action/replication_record"}},"created_at":"2026-08-05T01:37:16.959536+00:00","updated_at":"2026-08-05T01:37:16.959536+00:00"}