{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:DTXCIUIBCJQUMAAJKBQPOS6L2M","short_pith_number":"pith:DTXCIUIB","schema_version":"1.0","canonical_sha256":"1cee24510112614600095060f74bcbd307eb053657af59a344a1f045879026eb","source":{"kind":"arxiv","id":"2009.12812","version":3},"attestation_state":"computed","paper":{"title":"TernaryBERT: Distillation-aware Ultra-low Bit BERT","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Lifeng Shang, Lu Hou, Qun Liu, Wei Zhang, Xiao Chen, Xin Jiang, Yichun Yin","submitted_at":"2020-09-27T10:17:28Z","abstract_excerpt":"Transformer-based pre-training models like BERT have achieved remarkable performance in many natural language processing tasks.However, these models are both computation and memory expensive, hindering their deployment to resource-constrained devices. In this work, we propose TernaryBERT, which ternarizes the weights in a fine-tuned BERT model. Specifically, we use both approximation-based and loss-aware ternarization methods and empirically investigate the ternarization granularity of different parts of BERT. Moreover, to reduce the accuracy degradation caused by the lower capacity of low bit"},"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":"2009.12812","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-09-27T10:17:28Z","cross_cats_sorted":["cs.LG","cs.SD","eess.AS"],"title_canon_sha256":"26809c3415abca7a8fa24bc9b490de106f201f574726d33ac6461694272394bd","abstract_canon_sha256":"2d7327b3991df9ceeb664a2ee440fa19075c0b1eead56ab617e39a87da2dd0e8"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:41:57.494435Z","signature_b64":"CDIfdDcXNnUM/SFcaf1bD3ulSHXxTXi++rbzbbtdK66FM61w/SdINI7aOfKX45BFhln76DN7ROErcsFkMn3TAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1cee24510112614600095060f74bcbd307eb053657af59a344a1f045879026eb","last_reissued_at":"2026-07-05T01:41:57.493991Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:41:57.493991Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"TernaryBERT: Distillation-aware Ultra-low Bit BERT","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG","cs.SD","eess.AS"],"primary_cat":"cs.CL","authors_text":"Lifeng Shang, Lu Hou, Qun Liu, Wei Zhang, Xiao Chen, Xin Jiang, Yichun Yin","submitted_at":"2020-09-27T10:17:28Z","abstract_excerpt":"Transformer-based pre-training models like BERT have achieved remarkable performance in many natural language processing tasks.However, these models are both computation and memory expensive, hindering their deployment to resource-constrained devices. In this work, we propose TernaryBERT, which ternarizes the weights in a fine-tuned BERT model. Specifically, we use both approximation-based and loss-aware ternarization methods and empirically investigate the ternarization granularity of different parts of BERT. Moreover, to reduce the accuracy degradation caused by the lower capacity of low bit"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2009.12812","kind":"arxiv","version":3},"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/2009.12812/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":"2009.12812","created_at":"2026-07-05T01:41:57.494048+00:00"},{"alias_kind":"arxiv_version","alias_value":"2009.12812v3","created_at":"2026-07-05T01:41:57.494048+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2009.12812","created_at":"2026-07-05T01:41:57.494048+00:00"},{"alias_kind":"pith_short_12","alias_value":"DTXCIUIBCJQU","created_at":"2026-07-05T01:41:57.494048+00:00"},{"alias_kind":"pith_short_16","alias_value":"DTXCIUIBCJQUMAAJ","created_at":"2026-07-05T01:41:57.494048+00:00"},{"alias_kind":"pith_short_8","alias_value":"DTXCIUIB","created_at":"2026-07-05T01:41:57.494048+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.03957","citing_title":"BWTA: Accurate and Efficient Binarized Transformer by Algorithm-Hardware Co-design","ref_index":53,"is_internal_anchor":false},{"citing_arxiv_id":"2604.11321","citing_title":"Winner-Take-All Spiking Transformer for Language Modeling","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08028","citing_title":"A Comparative Study of Semantic Log Representations for Software Log-based Anomaly Detection","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04523","citing_title":"LOCALUT: Harnessing Capacity-Computation Tradeoffs for LUT-Based Inference in DRAM-PIM","ref_index":98,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12365","citing_title":"Adaptive Spiking Neurons for Vision and Language Modeling","ref_index":39,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M","json":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M.json","graph_json":"https://pith.science/api/pith-number/DTXCIUIBCJQUMAAJKBQPOS6L2M/graph.json","events_json":"https://pith.science/api/pith-number/DTXCIUIBCJQUMAAJKBQPOS6L2M/events.json","paper":"https://pith.science/paper/DTXCIUIB"},"agent_actions":{"view_html":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M","download_json":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M.json","view_paper":"https://pith.science/paper/DTXCIUIB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2009.12812&json=true","fetch_graph":"https://pith.science/api/pith-number/DTXCIUIBCJQUMAAJKBQPOS6L2M/graph.json","fetch_events":"https://pith.science/api/pith-number/DTXCIUIBCJQUMAAJKBQPOS6L2M/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M/action/timestamp_anchor","attest_storage":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M/action/storage_attestation","attest_author":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M/action/author_attestation","sign_citation":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M/action/citation_signature","submit_replication":"https://pith.science/pith/DTXCIUIBCJQUMAAJKBQPOS6L2M/action/replication_record"}},"created_at":"2026-07-05T01:41:57.494048+00:00","updated_at":"2026-07-05T01:41:57.494048+00:00"}