{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XHRWC26ITSP4B4W6ZQPYX5OBEO","short_pith_number":"pith:XHRWC26I","schema_version":"1.0","canonical_sha256":"b9e3616bc89c9fc0f2decc1f8bf5c123bc35f6be743fbd84b7f70147a50cbf1e","source":{"kind":"arxiv","id":"2412.18653","version":1},"attestation_state":"computed","paper":{"title":"1.58-bit FLUX","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Celong Liu, Chenglin Yang, Dongwon Kim, Liang-Chieh Chen, Xiaohui Shen, Xing Mei, Xueqing Deng","submitted_at":"2024-12-24T19:00:02Z","abstract_excerpt":"We present 1.58-bit FLUX, the first successful approach to quantizing the state-of-the-art text-to-image generation model, FLUX.1-dev, using 1.58-bit weights (i.e., values in {-1, 0, +1}) while maintaining comparable performance for generating 1024 x 1024 images. Notably, our quantization method operates without access to image data, relying solely on self-supervision from the FLUX.1-dev model. Additionally, we develop a custom kernel optimized for 1.58-bit operations, achieving a 7.7x reduction in model storage, a 5.1x reduction in inference memory, and improved inference latency. Extensive e"},"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":"2412.18653","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-12-24T19:00:02Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"a7f39abc1febe98935265a29ca8cd73d78979711fff4c8a0312d58ff3fc6a142","abstract_canon_sha256":"d911f3bf636aecc8583c88a4e8db6ec81568e090028e1f3596540314253d8e5c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T09:54:08.535419Z","signature_b64":"8jKrXFfi2miyQi0oYA8jRvHMZKoEzOMG3eUQOHSMpcK1Y/3BnpbkRoh7vLli5icLhn43TviW75JjnKmOOcVmAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b9e3616bc89c9fc0f2decc1f8bf5c123bc35f6be743fbd84b7f70147a50cbf1e","last_reissued_at":"2026-07-05T09:54:08.534983Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T09:54:08.534983Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"1.58-bit FLUX","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CV","authors_text":"Celong Liu, Chenglin Yang, Dongwon Kim, Liang-Chieh Chen, Xiaohui Shen, Xing Mei, Xueqing Deng","submitted_at":"2024-12-24T19:00:02Z","abstract_excerpt":"We present 1.58-bit FLUX, the first successful approach to quantizing the state-of-the-art text-to-image generation model, FLUX.1-dev, using 1.58-bit weights (i.e., values in {-1, 0, +1}) while maintaining comparable performance for generating 1024 x 1024 images. Notably, our quantization method operates without access to image data, relying solely on self-supervision from the FLUX.1-dev model. Additionally, we develop a custom kernel optimized for 1.58-bit operations, achieving a 7.7x reduction in model storage, a 5.1x reduction in inference memory, and improved inference latency. Extensive e"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2412.18653","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/2412.18653/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":"2412.18653","created_at":"2026-07-05T09:54:08.535044+00:00"},{"alias_kind":"arxiv_version","alias_value":"2412.18653v1","created_at":"2026-07-05T09:54:08.535044+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2412.18653","created_at":"2026-07-05T09:54:08.535044+00:00"},{"alias_kind":"pith_short_12","alias_value":"XHRWC26ITSP4","created_at":"2026-07-05T09:54:08.535044+00:00"},{"alias_kind":"pith_short_16","alias_value":"XHRWC26ITSP4B4W6","created_at":"2026-07-05T09:54:08.535044+00:00"},{"alias_kind":"pith_short_8","alias_value":"XHRWC26I","created_at":"2026-07-05T09:54:08.535044+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.00273","citing_title":"When Do Diffusion Models learn to Generate Multiple Objects?","ref_index":29,"is_internal_anchor":false},{"citing_arxiv_id":"2605.26632","citing_title":"RT-Lynx: Putting the GEMM Sparsity In a Right Way for Diffusion Models","ref_index":68,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04913","citing_title":"A Frame is Worth One Token: Efficient Generative World Modeling with Delta Tokens","ref_index":79,"is_internal_anchor":false},{"citing_arxiv_id":"2604.15521","citing_title":"Frequency-Aware Flow Matching for High-Quality Image Generation","ref_index":59,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO","json":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO.json","graph_json":"https://pith.science/api/pith-number/XHRWC26ITSP4B4W6ZQPYX5OBEO/graph.json","events_json":"https://pith.science/api/pith-number/XHRWC26ITSP4B4W6ZQPYX5OBEO/events.json","paper":"https://pith.science/paper/XHRWC26I"},"agent_actions":{"view_html":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO","download_json":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO.json","view_paper":"https://pith.science/paper/XHRWC26I","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2412.18653&json=true","fetch_graph":"https://pith.science/api/pith-number/XHRWC26ITSP4B4W6ZQPYX5OBEO/graph.json","fetch_events":"https://pith.science/api/pith-number/XHRWC26ITSP4B4W6ZQPYX5OBEO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO/action/storage_attestation","attest_author":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO/action/author_attestation","sign_citation":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO/action/citation_signature","submit_replication":"https://pith.science/pith/XHRWC26ITSP4B4W6ZQPYX5OBEO/action/replication_record"}},"created_at":"2026-07-05T09:54:08.535044+00:00","updated_at":"2026-07-05T09:54:08.535044+00:00"}