{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:34CWI64XHPEQOYNFVY7ATRM4XG","short_pith_number":"pith:34CWI64X","schema_version":"1.0","canonical_sha256":"df05647b973bc90761a5ae3e09c59cb998f9ea9488fe1d492a329be6b4021c7e","source":{"kind":"arxiv","id":"2504.14365","version":1},"attestation_state":"computed","paper":{"title":"Accelerating LLM Inference with Flexible N:M Sparsity via A Fully Digital Compute-in-Memory Accelerator","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.LG","authors_text":"Akshat Ramachandran, Arnab Raha, Deepak K. Mathaikutty, Shamik Kundu, Souvik Kundu, Tushar Krishna","submitted_at":"2025-04-19T17:47:01Z","abstract_excerpt":"Large language model (LLM) pruning with fixed N:M structured sparsity significantly limits the expressivity of the sparse model, yielding sub-optimal performance. In contrast, supporting multiple N:M patterns to provide sparse representational freedom introduces costly overhead in hardware. To address these challenges for LLMs, we first present a flexible layer-wise outlier-density-aware N:M sparsity (FLOW) selection method. FLOW enables the identification of optimal layer-wise N and M values (from a given range) by simultaneously accounting for the presence and distribution of outliers, allow"},"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":"2504.14365","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-04-19T17:47:01Z","cross_cats_sorted":["cs.AI","cs.AR"],"title_canon_sha256":"b7bacaa644e0db16b12802bbb6d80dbf1476400601c0f6228737c4993fed9a0a","abstract_canon_sha256":"53152d4c2c2e6e3283afb24196343687a9c39e13c250d9ecb7d3204c3449a1c3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:51:42.315081Z","signature_b64":"jBOvnbV1Ki6H5Ld6fVBOyfQr2hwZ7UgMSlGn3cTr4WuYyKlFYruItiDVNSJ18xaY73Ps7Sy4ltPE0g/NBmpvDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"df05647b973bc90761a5ae3e09c59cb998f9ea9488fe1d492a329be6b4021c7e","last_reissued_at":"2026-07-05T10:51:42.314586Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:51:42.314586Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Accelerating LLM Inference with Flexible N:M Sparsity via A Fully Digital Compute-in-Memory Accelerator","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.AR"],"primary_cat":"cs.LG","authors_text":"Akshat Ramachandran, Arnab Raha, Deepak K. Mathaikutty, Shamik Kundu, Souvik Kundu, Tushar Krishna","submitted_at":"2025-04-19T17:47:01Z","abstract_excerpt":"Large language model (LLM) pruning with fixed N:M structured sparsity significantly limits the expressivity of the sparse model, yielding sub-optimal performance. In contrast, supporting multiple N:M patterns to provide sparse representational freedom introduces costly overhead in hardware. To address these challenges for LLMs, we first present a flexible layer-wise outlier-density-aware N:M sparsity (FLOW) selection method. FLOW enables the identification of optimal layer-wise N and M values (from a given range) by simultaneously accounting for the presence and distribution of outliers, allow"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2504.14365","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/2504.14365/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":"2504.14365","created_at":"2026-07-05T10:51:42.314643+00:00"},{"alias_kind":"arxiv_version","alias_value":"2504.14365v1","created_at":"2026-07-05T10:51:42.314643+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2504.14365","created_at":"2026-07-05T10:51:42.314643+00:00"},{"alias_kind":"pith_short_12","alias_value":"34CWI64XHPEQ","created_at":"2026-07-05T10:51:42.314643+00:00"},{"alias_kind":"pith_short_16","alias_value":"34CWI64XHPEQOYNF","created_at":"2026-07-05T10:51:42.314643+00:00"},{"alias_kind":"pith_short_8","alias_value":"34CWI64X","created_at":"2026-07-05T10:51:42.314643+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/34CWI64XHPEQOYNFVY7ATRM4XG","json":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG.json","graph_json":"https://pith.science/api/pith-number/34CWI64XHPEQOYNFVY7ATRM4XG/graph.json","events_json":"https://pith.science/api/pith-number/34CWI64XHPEQOYNFVY7ATRM4XG/events.json","paper":"https://pith.science/paper/34CWI64X"},"agent_actions":{"view_html":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG","download_json":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG.json","view_paper":"https://pith.science/paper/34CWI64X","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2504.14365&json=true","fetch_graph":"https://pith.science/api/pith-number/34CWI64XHPEQOYNFVY7ATRM4XG/graph.json","fetch_events":"https://pith.science/api/pith-number/34CWI64XHPEQOYNFVY7ATRM4XG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG/action/storage_attestation","attest_author":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG/action/author_attestation","sign_citation":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG/action/citation_signature","submit_replication":"https://pith.science/pith/34CWI64XHPEQOYNFVY7ATRM4XG/action/replication_record"}},"created_at":"2026-07-05T10:51:42.314643+00:00","updated_at":"2026-07-05T10:51:42.314643+00:00"}