{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PCJVVNDK4BTKWV4AEFACDANWHN","short_pith_number":"pith:PCJVVNDK","schema_version":"1.0","canonical_sha256":"78935ab46ae066ab578021402181b63b547c4a729f3ca19e2d84ac6641228734","source":{"kind":"arxiv","id":"2502.05003","version":2},"attestation_state":"computed","paper":{"title":"QuEST: Stable Training of LLMs with 1-Bit Weights and Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andrei Panferov, Dan Alistarh, Jiale Chen, Mahdi Nikdan, Roberto L. Castro, Soroush Tabesh","submitted_at":"2025-02-07T15:23:34Z","abstract_excerpt":"One approach to reducing the massive costs of large language models (LLMs) is the use of quantized or sparse representations for training or deployment. While post-training compression methods are very popular, the question of obtaining even more accurate compressed models by directly training over such representations, i.e., Quantization-Aware Training (QAT), is still open: for example, a recent study (arXiv:2411.04330) put the \"optimal\" bit-width at which models can be trained using QAT, while staying accuracy-competitive with standard FP16/BF16 precision, at 8-bits weights and activations. "},"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":"2502.05003","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2025-02-07T15:23:34Z","cross_cats_sorted":[],"title_canon_sha256":"ecb9f9a28b73a8cf7a5920719538a83bf558714f7b3007b0ba373d7223733a47","abstract_canon_sha256":"4666b7a3d9ab6b316e934fcde7b4c917bb0ff4dcdf19937d0a7e4c840b9fa4c7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:19:24.355711Z","signature_b64":"lP7klJ0AOCoIRMDgXuRlNHcm9crmgXJWN511sk5KLfjKNaFSQUGvueOIkdunvnsQsYvZVrevDD+maWMyZwfPAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"78935ab46ae066ab578021402181b63b547c4a729f3ca19e2d84ac6641228734","last_reissued_at":"2026-07-05T11:19:24.355200Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:19:24.355200Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"QuEST: Stable Training of LLMs with 1-Bit Weights and Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Andrei Panferov, Dan Alistarh, Jiale Chen, Mahdi Nikdan, Roberto L. Castro, Soroush Tabesh","submitted_at":"2025-02-07T15:23:34Z","abstract_excerpt":"One approach to reducing the massive costs of large language models (LLMs) is the use of quantized or sparse representations for training or deployment. While post-training compression methods are very popular, the question of obtaining even more accurate compressed models by directly training over such representations, i.e., Quantization-Aware Training (QAT), is still open: for example, a recent study (arXiv:2411.04330) put the \"optimal\" bit-width at which models can be trained using QAT, while staying accuracy-competitive with standard FP16/BF16 precision, at 8-bits weights and activations. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.05003","kind":"arxiv","version":2},"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/2502.05003/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":"2502.05003","created_at":"2026-07-05T11:19:24.355264+00:00"},{"alias_kind":"arxiv_version","alias_value":"2502.05003v2","created_at":"2026-07-05T11:19:24.355264+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.05003","created_at":"2026-07-05T11:19:24.355264+00:00"},{"alias_kind":"pith_short_12","alias_value":"PCJVVNDK4BTK","created_at":"2026-07-05T11:19:24.355264+00:00"},{"alias_kind":"pith_short_16","alias_value":"PCJVVNDK4BTKWV4A","created_at":"2026-07-05T11:19:24.355264+00:00"},{"alias_kind":"pith_short_8","alias_value":"PCJVVNDK","created_at":"2026-07-05T11:19:24.355264+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.12327","citing_title":"Grid Games: The Power of Multiple Grids for Quantizing Large Language Models","ref_index":28,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN","json":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN.json","graph_json":"https://pith.science/api/pith-number/PCJVVNDK4BTKWV4AEFACDANWHN/graph.json","events_json":"https://pith.science/api/pith-number/PCJVVNDK4BTKWV4AEFACDANWHN/events.json","paper":"https://pith.science/paper/PCJVVNDK"},"agent_actions":{"view_html":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN","download_json":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN.json","view_paper":"https://pith.science/paper/PCJVVNDK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2502.05003&json=true","fetch_graph":"https://pith.science/api/pith-number/PCJVVNDK4BTKWV4AEFACDANWHN/graph.json","fetch_events":"https://pith.science/api/pith-number/PCJVVNDK4BTKWV4AEFACDANWHN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN/action/storage_attestation","attest_author":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN/action/author_attestation","sign_citation":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN/action/citation_signature","submit_replication":"https://pith.science/pith/PCJVVNDK4BTKWV4AEFACDANWHN/action/replication_record"}},"created_at":"2026-07-05T11:19:24.355264+00:00","updated_at":"2026-07-05T11:19:24.355264+00:00"}