{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:KGRULLYAMHOJIAQOMAO6YNPWSP","short_pith_number":"pith:KGRULLYA","schema_version":"1.0","canonical_sha256":"51a345af0061dc94020e601dec35f693cca38f03f2ea5cc670f66054d63d2e7b","source":{"kind":"arxiv","id":"2608.13426","version":1},"attestation_state":"computed","paper":{"title":"Reduced Matrix Multiplication: Input-Adaptive Matrix-Product Reduction for LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Jiawei Zhou, Yanhong Li, Zixuan Lan","submitted_at":"2026-08-13T16:16:04Z","abstract_excerpt":"Transformer-based language models achieve strong performance but incur substantial inference cost due to repeated high-dimensional matrix multiplications. We propose Reduced Matrix Multiplication (RMM), a training-free, input-adaptive inference method that reduces Transformer matrix products by selecting informative slices along their contraction dimensions, without modifying model weights. Under a simple retention-ratio control, RMM provides a smooth and predictable accuracy-efficiency trade-off. Across language models ranging from 1B to 70B parameters, we find that reduction tolerance depend"},"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":"2608.13426","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-08-13T16:16:04Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"f6a92ebdf8cb3c9e2f3b0cf0a6e428f43045720116fc8a03f2f117103369a301","abstract_canon_sha256":"452806c7a7d5c5fee8b972171a67656cebc3aa0eee2448a605d2da3f59d228fb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-14T01:04:25.110156Z","signature_b64":"jyy/ALNsgUlbRkNKbLqurN7dH7UcHXZOh3Ez8wUKWHUwrXf9vDQ30eA8MIOJjXQblQhDT+RW1fE314CkwTo8BA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"51a345af0061dc94020e601dec35f693cca38f03f2ea5cc670f66054d63d2e7b","last_reissued_at":"2026-08-14T01:04:25.108478Z","signature_status":"signed_v1","first_computed_at":"2026-08-14T01:04:25.108478Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Reduced Matrix Multiplication: Input-Adaptive Matrix-Product Reduction for LLM Inference","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Jiawei Zhou, Yanhong Li, Zixuan Lan","submitted_at":"2026-08-13T16:16:04Z","abstract_excerpt":"Transformer-based language models achieve strong performance but incur substantial inference cost due to repeated high-dimensional matrix multiplications. We propose Reduced Matrix Multiplication (RMM), a training-free, input-adaptive inference method that reduces Transformer matrix products by selecting informative slices along their contraction dimensions, without modifying model weights. Under a simple retention-ratio control, RMM provides a smooth and predictable accuracy-efficiency trade-off. Across language models ranging from 1B to 70B parameters, we find that reduction tolerance depend"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.13426","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/2608.13426/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":"2608.13426","created_at":"2026-08-14T01:04:25.109350+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.13426v1","created_at":"2026-08-14T01:04:25.109350+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.13426","created_at":"2026-08-14T01:04:25.109350+00:00"},{"alias_kind":"pith_short_12","alias_value":"KGRULLYAMHOJ","created_at":"2026-08-14T01:04:25.109350+00:00"},{"alias_kind":"pith_short_16","alias_value":"KGRULLYAMHOJIAQO","created_at":"2026-08-14T01:04:25.109350+00:00"},{"alias_kind":"pith_short_8","alias_value":"KGRULLYA","created_at":"2026-08-14T01:04:25.109350+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/KGRULLYAMHOJIAQOMAO6YNPWSP","json":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP.json","graph_json":"https://pith.science/api/pith-number/KGRULLYAMHOJIAQOMAO6YNPWSP/graph.json","events_json":"https://pith.science/api/pith-number/KGRULLYAMHOJIAQOMAO6YNPWSP/events.json","paper":"https://pith.science/paper/KGRULLYA"},"agent_actions":{"view_html":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP","download_json":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP.json","view_paper":"https://pith.science/paper/KGRULLYA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.13426&json=true","fetch_graph":"https://pith.science/api/pith-number/KGRULLYAMHOJIAQOMAO6YNPWSP/graph.json","fetch_events":"https://pith.science/api/pith-number/KGRULLYAMHOJIAQOMAO6YNPWSP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP/action/storage_attestation","attest_author":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP/action/author_attestation","sign_citation":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP/action/citation_signature","submit_replication":"https://pith.science/pith/KGRULLYAMHOJIAQOMAO6YNPWSP/action/replication_record"}},"created_at":"2026-08-14T01:04:25.109350+00:00","updated_at":"2026-08-14T01:04:25.109350+00:00"}