{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:E7IU2HSFRM3JPKCL2DVD5NHY2T","short_pith_number":"pith:E7IU2HSF","schema_version":"1.0","canonical_sha256":"27d14d1e458b3697a84bd0ea3eb4f8d4e723bb36044cebdc09f3002a11c75087","source":{"kind":"arxiv","id":"2412.00648","version":4},"attestation_state":"computed","paper":{"title":"DFRot: Achieving Outlier-Free and Massive Activation-Free for Rotated LLMs with Refined Rotation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Jingyang Xiang, Sai Qian Zhang","submitted_at":"2024-12-01T02:55:08Z","abstract_excerpt":"Rotating the activation and weight matrices to reduce the influence of outliers in large language models (LLMs) has recently attracted significant attention, particularly in the context of model quantization. Prior studies have shown that in low-precision quantization scenarios, such as 4-bit weights and 4-bit activations (W4A4), randomized Hadamard transforms can achieve significantly higher accuracy than randomized orthogonal transforms. Notably, the reason behind this phenomenon remains unknown. In this paper, we find that these transformations show substantial improvement in eliminating ou"},"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":"2412.00648","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2024-12-01T02:55:08Z","cross_cats_sorted":["stat.ML"],"title_canon_sha256":"66ff68b4c60e6354bf8665a493a011c495ee486f76b799c311ba0d6855ffafc5","abstract_canon_sha256":"3858700839a92d45194c40c6132170682bacee83a7dc598d515dc631e68fcb53"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:37:12.666186Z","signature_b64":"Kr/x32oMZX6fXOBwwVfbzar9G0xaaUT1+YF857ZeadI6C1PJnYR9gnfiCqo1aiVWZ2nDDSPuFvOmW1HAkGooDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"27d14d1e458b3697a84bd0ea3eb4f8d4e723bb36044cebdc09f3002a11c75087","last_reissued_at":"2026-07-05T11:37:12.665683Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:37:12.665683Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"DFRot: Achieving Outlier-Free and Massive Activation-Free for Rotated LLMs with Refined Rotation","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["stat.ML"],"primary_cat":"cs.LG","authors_text":"Jingyang Xiang, Sai Qian Zhang","submitted_at":"2024-12-01T02:55:08Z","abstract_excerpt":"Rotating the activation and weight matrices to reduce the influence of outliers in large language models (LLMs) has recently attracted significant attention, particularly in the context of model quantization. Prior studies have shown that in low-precision quantization scenarios, such as 4-bit weights and 4-bit activations (W4A4), randomized Hadamard transforms can achieve significantly higher accuracy than randomized orthogonal transforms. Notably, the reason behind this phenomenon remains unknown. In this paper, we find that these transformations show substantial improvement in eliminating ou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.00648","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/2412.00648/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":"2412.00648","created_at":"2026-07-05T11:37:12.665748+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.00648v4","created_at":"2026-07-05T11:37:12.665748+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.00648","created_at":"2026-07-05T11:37:12.665748+00:00"},{"alias_kind":"pith_short_12","alias_value":"E7IU2HSFRM3J","created_at":"2026-07-05T11:37:12.665748+00:00"},{"alias_kind":"pith_short_16","alias_value":"E7IU2HSFRM3JPKCL","created_at":"2026-07-05T11:37:12.665748+00:00"},{"alias_kind":"pith_short_8","alias_value":"E7IU2HSF","created_at":"2026-07-05T11:37:12.665748+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.00573","citing_title":"LASER: Loss-Aware Singular-value Decomposition and Rank Allocation for Efficient Low-Precision Vision-Language Models","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02570","citing_title":"WSVD: Weighted Low-Rank Approximation for Fast and Efficient Execution of Low-Precision Vision-Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04013","citing_title":"RUQuant: Towards Refining Uniform Quantization for Large Language Models","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T","json":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T.json","graph_json":"https://pith.science/api/pith-number/E7IU2HSFRM3JPKCL2DVD5NHY2T/graph.json","events_json":"https://pith.science/api/pith-number/E7IU2HSFRM3JPKCL2DVD5NHY2T/events.json","paper":"https://pith.science/paper/E7IU2HSF"},"agent_actions":{"view_html":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T","download_json":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T.json","view_paper":"https://pith.science/paper/E7IU2HSF","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.00648&json=true","fetch_graph":"https://pith.science/api/pith-number/E7IU2HSFRM3JPKCL2DVD5NHY2T/graph.json","fetch_events":"https://pith.science/api/pith-number/E7IU2HSFRM3JPKCL2DVD5NHY2T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T/action/storage_attestation","attest_author":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T/action/author_attestation","sign_citation":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T/action/citation_signature","submit_replication":"https://pith.science/pith/E7IU2HSFRM3JPKCL2DVD5NHY2T/action/replication_record"}},"created_at":"2026-07-05T11:37:12.665748+00:00","updated_at":"2026-07-05T11:37:12.665748+00:00"}