{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:IUBUP3YA3XPNKTMQHJP5TA4IBR","short_pith_number":"pith:IUBUP3YA","schema_version":"1.0","canonical_sha256":"450347ef00ddded54d903a5fd983880c54d424ae58ef62951c9f890dd7013d09","source":{"kind":"arxiv","id":"2309.15531","version":4},"attestation_state":"computed","paper":{"title":"Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Beomseok Kwon, Byeongwook Kim, Dongsoo Lee, Jeonghoon Kim, Jung Hwan Heo, Se Jung Kwon","submitted_at":"2023-09-27T09:48:31Z","abstract_excerpt":"Large Language Models (LLMs) have recently demonstrated remarkable success across various tasks. However, efficiently serving LLMs has been a challenge due to the large memory bottleneck, specifically in small batch inference settings (e.g. mobile devices). Weight-only quantization can be a promising approach, but sub-4 bit quantization remains a challenge due to large-magnitude activation outliers. To mitigate the undesirable outlier effect, we first propose per-IC quantization, a simple yet effective method that creates quantization groups within each input channel (IC) rather than the conve"},"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":"2309.15531","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2023-09-27T09:48:31Z","cross_cats_sorted":[],"title_canon_sha256":"a2c24b65c0adedd0c59fc8d4df4d77a2277aec6a6411cd359de5c2555965dee4","abstract_canon_sha256":"29f7b2bef0c3f946aeef33e575d61f982976f7bf3c214ca9ec9d26502872a76f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:48:13.037780Z","signature_b64":"ypyqQfLafQib585k4cR0uP2xWFNOQ132mp8j64bm7GgttDysIlnGB9ihwoWLzd2560qJAtVVMFxsJL7zau5bBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"450347ef00ddded54d903a5fd983880c54d424ae58ef62951c9f890dd7013d09","last_reissued_at":"2026-07-05T10:48:13.037298Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:48:13.037298Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Beomseok Kwon, Byeongwook Kim, Dongsoo Lee, Jeonghoon Kim, Jung Hwan Heo, Se Jung Kwon","submitted_at":"2023-09-27T09:48:31Z","abstract_excerpt":"Large Language Models (LLMs) have recently demonstrated remarkable success across various tasks. However, efficiently serving LLMs has been a challenge due to the large memory bottleneck, specifically in small batch inference settings (e.g. mobile devices). Weight-only quantization can be a promising approach, but sub-4 bit quantization remains a challenge due to large-magnitude activation outliers. To mitigate the undesirable outlier effect, we first propose per-IC quantization, a simple yet effective method that creates quantization groups within each input channel (IC) rather than the conve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2309.15531","kind":"arxiv","version":4},"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/2309.15531/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":"2309.15531","created_at":"2026-07-05T10:48:13.037361+00:00"},{"alias_kind":"arxiv_version","alias_value":"2309.15531v4","created_at":"2026-07-05T10:48:13.037361+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2309.15531","created_at":"2026-07-05T10:48:13.037361+00:00"},{"alias_kind":"pith_short_12","alias_value":"IUBUP3YA3XPN","created_at":"2026-07-05T10:48:13.037361+00:00"},{"alias_kind":"pith_short_16","alias_value":"IUBUP3YA3XPNKTMQ","created_at":"2026-07-05T10:48:13.037361+00:00"},{"alias_kind":"pith_short_8","alias_value":"IUBUP3YA","created_at":"2026-07-05T10:48:13.037361+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26587","citing_title":"SharQ: Bridging Activation Sparsity and FP4 Quantization for LLM Inference","ref_index":79,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR","json":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR.json","graph_json":"https://pith.science/api/pith-number/IUBUP3YA3XPNKTMQHJP5TA4IBR/graph.json","events_json":"https://pith.science/api/pith-number/IUBUP3YA3XPNKTMQHJP5TA4IBR/events.json","paper":"https://pith.science/paper/IUBUP3YA"},"agent_actions":{"view_html":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR","download_json":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR.json","view_paper":"https://pith.science/paper/IUBUP3YA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2309.15531&json=true","fetch_graph":"https://pith.science/api/pith-number/IUBUP3YA3XPNKTMQHJP5TA4IBR/graph.json","fetch_events":"https://pith.science/api/pith-number/IUBUP3YA3XPNKTMQHJP5TA4IBR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR/action/storage_attestation","attest_author":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR/action/author_attestation","sign_citation":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR/action/citation_signature","submit_replication":"https://pith.science/pith/IUBUP3YA3XPNKTMQHJP5TA4IBR/action/replication_record"}},"created_at":"2026-07-05T10:48:13.037361+00:00","updated_at":"2026-07-05T10:48:13.037361+00:00"}