{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FCZGIG6YM6MT5QELDJIKM7ZAA6","short_pith_number":"pith:FCZGIG6Y","schema_version":"1.0","canonical_sha256":"28b2641bd867993ec08b1a50a67f20079be979139708f2d2798dec5064f538c1","source":{"kind":"arxiv","id":"2603.15854","version":2},"attestation_state":"computed","paper":{"title":"FlashSampling: Fast and Memory-Efficient Exact Sampling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"FlashSampling fuses exact categorical sampling into the LM-head matrix multiply so the full logits tensor is never written to HBM.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Mengdi Wang, Tomas Ruiz, Xuyang Shen, Yifan Zhang, Yiran Zhong, Zhen Qin","submitted_at":"2026-03-16T19:37:08Z","abstract_excerpt":"Sampling from a categorical distribution is mathematically simple, but in large-vocabulary decoding, it often triggers extra memory traffic and extra kernels after the LM head. We present FlashSampling, an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in HBM. The method is simple: compute logits tile-by-tile on chip, add Gumbel noise, keep only one maximizer per row and per vocabulary tile, and finish with a small reduction over tiles. In tensor-parallel decoding, FlashSampling replaces the all-gather of logits with streaming peer"},"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":"2603.15854","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-03-16T19:37:08Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"e435c8655d84f9c3aa0afc24a8b8288f6ba2cdf4d13ca277893c96bab91d73a8","abstract_canon_sha256":"c0d3d84724b127fce2caceeb33a5adc2fc84b887648672cb76874fab169bb4a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T03:10:03.436072Z","signature_b64":"jpjY0Qsapxu/7gKmGKs5H4E5u2lcprvJBNdL+CcL6noCRZPTDrK+bd7R/IathgPywCGXEGNAYCvp1rRPHHFVAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"28b2641bd867993ec08b1a50a67f20079be979139708f2d2798dec5064f538c1","last_reissued_at":"2026-05-18T03:10:03.435468Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T03:10:03.435468Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FlashSampling: Fast and Memory-Efficient Exact Sampling","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"FlashSampling fuses exact categorical sampling into the LM-head matrix multiply so the full logits tensor is never written to HBM.","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Mengdi Wang, Tomas Ruiz, Xuyang Shen, Yifan Zhang, Yiran Zhong, Zhen Qin","submitted_at":"2026-03-16T19:37:08Z","abstract_excerpt":"Sampling from a categorical distribution is mathematically simple, but in large-vocabulary decoding, it often triggers extra memory traffic and extra kernels after the LM head. We present FlashSampling, an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in HBM. The method is simple: compute logits tile-by-tile on chip, add Gumbel noise, keep only one maximizer per row and per vocabulary tile, and finish with a small reduction over tiles. In tensor-parallel decoding, FlashSampling replaces the all-gather of logits with streaming peer"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"FlashSampling is an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in HBM; in tensor-parallel decoding it replaces all-gather with streaming peer-to-peer writes, achieving kernel-level speedups and up to 10% reduction in time per output token.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The claim that argmax decomposes exactly over vocabulary partitions (and that the hierarchical factorization for grouped variants preserves exact categorical sampling) holds without numerical error or edge-case failure when tiles are processed independently on-chip.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"FlashSampling performs exact Gumbel-max sampling inside the LM-head matmul via on-chip tiling and hierarchical argmax reduction, delivering up to 10% faster token generation in vLLM on datacenter GPUs without approximation.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"FlashSampling fuses exact categorical sampling into the LM-head matrix multiply so the full logits tensor is never written to HBM.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"ce3cba47136aac04afdbdcb5d0d61f83fc2176d9935ac901de94d3750e91ac7e"},"source":{"id":"2603.15854","kind":"arxiv","version":2},"verdict":{"id":"3f38b0a9-aca2-46ac-a984-5ab58db9a9e7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T09:57:45.197851Z","strongest_claim":"FlashSampling is an exact sampling primitive that fuses sampling into the LM-head matmul and never materializes the logits tensor in HBM; in tensor-parallel decoding it replaces all-gather with streaming peer-to-peer writes, achieving kernel-level speedups and up to 10% reduction in time per output token.","one_line_summary":"FlashSampling performs exact Gumbel-max sampling inside the LM-head matmul via on-chip tiling and hierarchical argmax reduction, delivering up to 10% faster token generation in vLLM on datacenter GPUs without approximation.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The claim that argmax decomposes exactly over vocabulary partitions (and that the hierarchical factorization for grouped variants preserves exact categorical sampling) holds without numerical error or edge-case failure when tiles are processed independently on-chip.","pith_extraction_headline":"FlashSampling fuses exact categorical sampling into the LM-head matrix multiply so the full logits tensor is never written to HBM."},"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":"2603.15854","created_at":"2026-05-18T03:10:03.435590+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.15854v2","created_at":"2026-05-18T03:10:03.435590+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.15854","created_at":"2026-05-18T03:10:03.435590+00:00"},{"alias_kind":"pith_short_12","alias_value":"FCZGIG6YM6MT","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_16","alias_value":"FCZGIG6YM6MT5QEL","created_at":"2026-05-18T12:33:37.589309+00:00"},{"alias_kind":"pith_short_8","alias_value":"FCZGIG6Y","created_at":"2026-05-18T12:33:37.589309+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2604.17384","citing_title":"Towards a Data-Parameter Correspondence for LLMs: A Preliminary Discussion","ref_index":40,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6","json":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6.json","graph_json":"https://pith.science/api/pith-number/FCZGIG6YM6MT5QELDJIKM7ZAA6/graph.json","events_json":"https://pith.science/api/pith-number/FCZGIG6YM6MT5QELDJIKM7ZAA6/events.json","paper":"https://pith.science/paper/FCZGIG6Y"},"agent_actions":{"view_html":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6","download_json":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6.json","view_paper":"https://pith.science/paper/FCZGIG6Y","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.15854&json=true","fetch_graph":"https://pith.science/api/pith-number/FCZGIG6YM6MT5QELDJIKM7ZAA6/graph.json","fetch_events":"https://pith.science/api/pith-number/FCZGIG6YM6MT5QELDJIKM7ZAA6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6/action/storage_attestation","attest_author":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6/action/author_attestation","sign_citation":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6/action/citation_signature","submit_replication":"https://pith.science/pith/FCZGIG6YM6MT5QELDJIKM7ZAA6/action/replication_record"}},"created_at":"2026-05-18T03:10:03.435590+00:00","updated_at":"2026-05-18T03:10:03.435590+00:00"}