{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:5ATZIG23KIGK37TA2QU34CSJGU","short_pith_number":"pith:5ATZIG23","schema_version":"1.0","canonical_sha256":"e827941b5b520cadfe60d429be0a4935239ca4ec9da9507b7f542c330fda9428","source":{"kind":"arxiv","id":"2501.01144","version":5},"attestation_state":"computed","paper":{"title":"BlockDialect: Block-wise Fine-grained Mixed Format Quantization for Energy-Efficient LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Thierry Tambe, Wonsuk Jang","submitted_at":"2025-01-02T08:57:00Z","abstract_excerpt":"The rapidly increasing size of large language models (LLMs) presents significant challenges in memory usage and computational costs. Quantizing both weights and activations can address these issues, with hardware-supported fine-grained scaling emerging as a promising solution to mitigate outliers. However, existing methods struggle to capture nuanced block data distributions. We propose BlockDialect, a block-wise fine-grained mixed format technique that assigns a per-block optimal number format from a formatbook for better data representation. Additionally, we introduce DialectFP4, a formatboo"},"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":"2501.01144","kind":"arxiv","version":5},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-02T08:57:00Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2d3e7295411dac0a59f35ecdf408d5406054c0395447a9e913b77b112d0085a8","abstract_canon_sha256":"7637836c260d8113a1f8cadfa66bf6f2d7aeff2495efd0fc9c0af13a7c0ebfc7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:42:24.957710Z","signature_b64":"N4ro5cP0cNviZJItMfrIt42K1oJtrWN8V+HeSgilGxnm/1tUSNoaqQ1CIDtmct1b6LLJQ0fgSue8pl2xMelSAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"e827941b5b520cadfe60d429be0a4935239ca4ec9da9507b7f542c330fda9428","last_reissued_at":"2026-07-05T11:42:24.957231Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:42:24.957231Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BlockDialect: Block-wise Fine-grained Mixed Format Quantization for Energy-Efficient LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Thierry Tambe, Wonsuk Jang","submitted_at":"2025-01-02T08:57:00Z","abstract_excerpt":"The rapidly increasing size of large language models (LLMs) presents significant challenges in memory usage and computational costs. Quantizing both weights and activations can address these issues, with hardware-supported fine-grained scaling emerging as a promising solution to mitigate outliers. However, existing methods struggle to capture nuanced block data distributions. We propose BlockDialect, a block-wise fine-grained mixed format technique that assigns a per-block optimal number format from a formatbook for better data representation. Additionally, we introduce DialectFP4, a formatboo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.01144","kind":"arxiv","version":5},"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/2501.01144/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":"2501.01144","created_at":"2026-07-05T11:42:24.957287+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.01144v5","created_at":"2026-07-05T11:42:24.957287+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.01144","created_at":"2026-07-05T11:42:24.957287+00:00"},{"alias_kind":"pith_short_12","alias_value":"5ATZIG23KIGK","created_at":"2026-07-05T11:42:24.957287+00:00"},{"alias_kind":"pith_short_16","alias_value":"5ATZIG23KIGK37TA","created_at":"2026-07-05T11:42:24.957287+00:00"},{"alias_kind":"pith_short_8","alias_value":"5ATZIG23","created_at":"2026-07-05T11:42:24.957287+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00365","citing_title":"SPARQLe: Sub-Precision Activation Representation for Quantized LLM Inference","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU","json":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU.json","graph_json":"https://pith.science/api/pith-number/5ATZIG23KIGK37TA2QU34CSJGU/graph.json","events_json":"https://pith.science/api/pith-number/5ATZIG23KIGK37TA2QU34CSJGU/events.json","paper":"https://pith.science/paper/5ATZIG23"},"agent_actions":{"view_html":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU","download_json":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU.json","view_paper":"https://pith.science/paper/5ATZIG23","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.01144&json=true","fetch_graph":"https://pith.science/api/pith-number/5ATZIG23KIGK37TA2QU34CSJGU/graph.json","fetch_events":"https://pith.science/api/pith-number/5ATZIG23KIGK37TA2QU34CSJGU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU/action/storage_attestation","attest_author":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU/action/author_attestation","sign_citation":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU/action/citation_signature","submit_replication":"https://pith.science/pith/5ATZIG23KIGK37TA2QU34CSJGU/action/replication_record"}},"created_at":"2026-07-05T11:42:24.957287+00:00","updated_at":"2026-07-05T11:42:24.957287+00:00"}