{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:GJWVK4GOIS4VHYYXPTH2XUEK7P","short_pith_number":"pith:GJWVK4GO","schema_version":"1.0","canonical_sha256":"326d5570ce44b953e3177ccfabd08afbfe40e7a268d6775af84ba83969a4d6d2","source":{"kind":"arxiv","id":"2004.02178","version":2},"attestation_state":"computed","paper":{"title":"FastBERT: a Self-distilling BERT with Adaptive Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Haotang Deng, Peng Zhou, Qi Ju, Weijie Liu, Zhe Zhao, Zhiruo Wang","submitted_at":"2020-04-05T12:29:20Z","abstract_excerpt":"Pre-trained language models like BERT have proven to be highly performant. However, they are often computationally expensive in many practical scenarios, for such heavy models can hardly be readily implemented with limited resources. To improve their efficiency with an assured model performance, we propose a novel speed-tunable FastBERT with adaptive inference time. The speed at inference can be flexibly adjusted under varying demands, while redundant calculation of samples is avoided. Moreover, this model adopts a unique self-distillation mechanism at fine-tuning, further enabling a greater c"},"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":"2004.02178","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-04-05T12:29:20Z","cross_cats_sorted":[],"title_canon_sha256":"60a1e8ce48f2d94486bafb013fbf9f76bb0cb05dd92e7688c2917bea0f23a8d7","abstract_canon_sha256":"17bc98c2c6b38719fd8351720d3f86f2b48ea3b91aa08616e47d04ec3176211e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T00:59:14.802269Z","signature_b64":"eEW1Ria6yvY6wsdxxubAZsgq6xhZlwl8aliLCrTdWRO83X7S6HZRZzjOpZDK8tVpJWjoxDiKi3Z/ObrnnKQJBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"326d5570ce44b953e3177ccfabd08afbfe40e7a268d6775af84ba83969a4d6d2","last_reissued_at":"2026-07-05T00:59:14.801798Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T00:59:14.801798Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"FastBERT: a Self-distilling BERT with Adaptive Inference Time","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Haotang Deng, Peng Zhou, Qi Ju, Weijie Liu, Zhe Zhao, Zhiruo Wang","submitted_at":"2020-04-05T12:29:20Z","abstract_excerpt":"Pre-trained language models like BERT have proven to be highly performant. However, they are often computationally expensive in many practical scenarios, for such heavy models can hardly be readily implemented with limited resources. To improve their efficiency with an assured model performance, we propose a novel speed-tunable FastBERT with adaptive inference time. The speed at inference can be flexibly adjusted under varying demands, while redundant calculation of samples is avoided. Moreover, this model adopts a unique self-distillation mechanism at fine-tuning, further enabling a greater c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2004.02178","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/2004.02178/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":"2004.02178","created_at":"2026-07-05T00:59:14.801856+00:00"},{"alias_kind":"arxiv_version","alias_value":"2004.02178v2","created_at":"2026-07-05T00:59:14.801856+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2004.02178","created_at":"2026-07-05T00:59:14.801856+00:00"},{"alias_kind":"pith_short_12","alias_value":"GJWVK4GOIS4V","created_at":"2026-07-05T00:59:14.801856+00:00"},{"alias_kind":"pith_short_16","alias_value":"GJWVK4GOIS4VHYYX","created_at":"2026-07-05T00:59:14.801856+00:00"},{"alias_kind":"pith_short_8","alias_value":"GJWVK4GO","created_at":"2026-07-05T00:59:14.801856+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.26538","citing_title":"CascadeFormer: Depth-Tapered Transformers Motivated by Gradient Fan-in Asymmetry","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2604.18592","citing_title":"Two-dimensional early exit optimisation of LLM inference","ref_index":17,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P","json":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P.json","graph_json":"https://pith.science/api/pith-number/GJWVK4GOIS4VHYYXPTH2XUEK7P/graph.json","events_json":"https://pith.science/api/pith-number/GJWVK4GOIS4VHYYXPTH2XUEK7P/events.json","paper":"https://pith.science/paper/GJWVK4GO"},"agent_actions":{"view_html":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P","download_json":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P.json","view_paper":"https://pith.science/paper/GJWVK4GO","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2004.02178&json=true","fetch_graph":"https://pith.science/api/pith-number/GJWVK4GOIS4VHYYXPTH2XUEK7P/graph.json","fetch_events":"https://pith.science/api/pith-number/GJWVK4GOIS4VHYYXPTH2XUEK7P/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P/action/storage_attestation","attest_author":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P/action/author_attestation","sign_citation":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P/action/citation_signature","submit_replication":"https://pith.science/pith/GJWVK4GOIS4VHYYXPTH2XUEK7P/action/replication_record"}},"created_at":"2026-07-05T00:59:14.801856+00:00","updated_at":"2026-07-05T00:59:14.801856+00:00"}