{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:AXFZ2C26VWI6OPHABQLQJD4TAE","short_pith_number":"pith:AXFZ2C26","schema_version":"1.0","canonical_sha256":"05cb9d0b5ead91e73ce00c17048f930130924f38d06364be10fa4858b4d05670","source":{"kind":"arxiv","id":"2203.13483","version":1},"attestation_state":"computed","paper":{"title":"MKQ-BERT: Quantized BERT with 4-bits Weights and Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hanlin Tang, Jianchen Zhu, Kai Liu, Xipeng Zhang, Zhanhui Kang","submitted_at":"2022-03-25T07:27:18Z","abstract_excerpt":"Recently, pre-trained Transformer based language models, such as BERT, have shown great superiority over the traditional methods in many Natural Language Processing (NLP) tasks. However, the computational cost for deploying these models is prohibitive on resource-restricted devices. One method to alleviate this computation overhead is to quantize the original model into fewer bits representation, and previous work has proved that we can at most quantize both weights and activations of BERT into 8-bits, without degrading its performance. In this work, we propose MKQ-BERT, which further improves"},"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":"2203.13483","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2022-03-25T07:27:18Z","cross_cats_sorted":[],"title_canon_sha256":"448e8a31cac726a0c1c3e639adfd18de448e42ca6e24a18c318ced082eb50fa6","abstract_canon_sha256":"87812a54796e59ab0a4a4e487f52d60580e098ec2aa6795c421682213b25b69f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:08:32.590642Z","signature_b64":"BOAVRG0Zn6EaTBpcQN1luMw7KZ1gbUsTG/LzrBGfmsNg2TKho4iMK6BSbW4dhWFMi5hbYr0hqq3GCVMIIXAsAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"05cb9d0b5ead91e73ce00c17048f930130924f38d06364be10fa4858b4d05670","last_reissued_at":"2026-07-05T04:08:32.590225Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:08:32.590225Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"MKQ-BERT: Quantized BERT with 4-bits Weights and Activations","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hanlin Tang, Jianchen Zhu, Kai Liu, Xipeng Zhang, Zhanhui Kang","submitted_at":"2022-03-25T07:27:18Z","abstract_excerpt":"Recently, pre-trained Transformer based language models, such as BERT, have shown great superiority over the traditional methods in many Natural Language Processing (NLP) tasks. However, the computational cost for deploying these models is prohibitive on resource-restricted devices. One method to alleviate this computation overhead is to quantize the original model into fewer bits representation, and previous work has proved that we can at most quantize both weights and activations of BERT into 8-bits, without degrading its performance. In this work, we propose MKQ-BERT, which further improves"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2203.13483","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/2203.13483/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":"2203.13483","created_at":"2026-07-05T04:08:32.590278+00:00"},{"alias_kind":"arxiv_version","alias_value":"2203.13483v1","created_at":"2026-07-05T04:08:32.590278+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2203.13483","created_at":"2026-07-05T04:08:32.590278+00:00"},{"alias_kind":"pith_short_12","alias_value":"AXFZ2C26VWI6","created_at":"2026-07-05T04:08:32.590278+00:00"},{"alias_kind":"pith_short_16","alias_value":"AXFZ2C26VWI6OPHA","created_at":"2026-07-05T04:08:32.590278+00:00"},{"alias_kind":"pith_short_8","alias_value":"AXFZ2C26","created_at":"2026-07-05T04:08:32.590278+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.00347","citing_title":"Pushing the Limits of Low-Bit Optimizers: A Focus on EMA Dynamics","ref_index":17,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE","json":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE.json","graph_json":"https://pith.science/api/pith-number/AXFZ2C26VWI6OPHABQLQJD4TAE/graph.json","events_json":"https://pith.science/api/pith-number/AXFZ2C26VWI6OPHABQLQJD4TAE/events.json","paper":"https://pith.science/paper/AXFZ2C26"},"agent_actions":{"view_html":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE","download_json":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE.json","view_paper":"https://pith.science/paper/AXFZ2C26","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2203.13483&json=true","fetch_graph":"https://pith.science/api/pith-number/AXFZ2C26VWI6OPHABQLQJD4TAE/graph.json","fetch_events":"https://pith.science/api/pith-number/AXFZ2C26VWI6OPHABQLQJD4TAE/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE/action/timestamp_anchor","attest_storage":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE/action/storage_attestation","attest_author":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE/action/author_attestation","sign_citation":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE/action/citation_signature","submit_replication":"https://pith.science/pith/AXFZ2C26VWI6OPHABQLQJD4TAE/action/replication_record"}},"created_at":"2026-07-05T04:08:32.590278+00:00","updated_at":"2026-07-05T04:08:32.590278+00:00"}