{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2020:5STRQ7LRMSQ74JFQL2EJDHFQVT","short_pith_number":"pith:5STRQ7LR","schema_version":"1.0","canonical_sha256":"eca7187d7164a1fe24b05e88919cb0acda744b8f2d682359f3270012e32da73a","source":{"kind":"arxiv","id":"2006.04152","version":3},"attestation_state":"computed","paper":{"title":"BERT Loses Patience: Fast and Robust Inference with Early Exit","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Canwen Xu, Furu Wei, Julian McAuley, Ke Xu, Tao Ge, Wangchunshu Zhou","submitted_at":"2020-06-07T13:38:32Z","abstract_excerpt":"In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimen"},"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":"2006.04152","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2020-06-07T13:38:32Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"f325821cef73a0df9c76064c27b80b7cb7a656eb6714ba34a985d26dbcb67c42","abstract_canon_sha256":"31fc0dce7100588d7cbaf5af58620940af03a7ea7f990872c2b86a07c76f929d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T01:45:03.928016Z","signature_b64":"s5cGrTl/IMhFHhirHQRR4ZEwJragBtDBh6uJSvtWBMkeQ2lMqIlS0QfpC1/Rn5uFCtrCr6Lq4uDKBrspeENrCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eca7187d7164a1fe24b05e88919cb0acda744b8f2d682359f3270012e32da73a","last_reissued_at":"2026-07-05T01:45:03.927537Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T01:45:03.927537Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BERT Loses Patience: Fast and Robust Inference with Early Exit","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Canwen Xu, Furu Wei, Julian McAuley, Ke Xu, Tao Ge, Wangchunshu Zhou","submitted_at":"2020-06-07T13:38:32Z","abstract_excerpt":"In this paper, we propose Patience-based Early Exit, a straightforward yet effective inference method that can be used as a plug-and-play technique to simultaneously improve the efficiency and robustness of a pretrained language model (PLM). To achieve this, our approach couples an internal-classifier with each layer of a PLM and dynamically stops inference when the intermediate predictions of the internal classifiers remain unchanged for a pre-defined number of steps. Our approach improves inference efficiency as it allows the model to make a prediction with fewer layers. Meanwhile, experimen"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2006.04152","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/2006.04152/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":"2006.04152","created_at":"2026-07-05T01:45:03.927595+00:00"},{"alias_kind":"arxiv_version","alias_value":"2006.04152v3","created_at":"2026-07-05T01:45:03.927595+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2006.04152","created_at":"2026-07-05T01:45:03.927595+00:00"},{"alias_kind":"pith_short_12","alias_value":"5STRQ7LRMSQ7","created_at":"2026-07-05T01:45:03.927595+00:00"},{"alias_kind":"pith_short_16","alias_value":"5STRQ7LRMSQ74JFQ","created_at":"2026-07-05T01:45:03.927595+00:00"},{"alias_kind":"pith_short_8","alias_value":"5STRQ7LR","created_at":"2026-07-05T01:45:03.927595+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.05527","citing_title":"EdgeServing: Deadline-Aware Multi-DNN Serving at the Edge","ref_index":13,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT","json":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT.json","graph_json":"https://pith.science/api/pith-number/5STRQ7LRMSQ74JFQL2EJDHFQVT/graph.json","events_json":"https://pith.science/api/pith-number/5STRQ7LRMSQ74JFQL2EJDHFQVT/events.json","paper":"https://pith.science/paper/5STRQ7LR"},"agent_actions":{"view_html":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT","download_json":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT.json","view_paper":"https://pith.science/paper/5STRQ7LR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2006.04152&json=true","fetch_graph":"https://pith.science/api/pith-number/5STRQ7LRMSQ74JFQL2EJDHFQVT/graph.json","fetch_events":"https://pith.science/api/pith-number/5STRQ7LRMSQ74JFQL2EJDHFQVT/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT/action/storage_attestation","attest_author":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT/action/author_attestation","sign_citation":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT/action/citation_signature","submit_replication":"https://pith.science/pith/5STRQ7LRMSQ74JFQL2EJDHFQVT/action/replication_record"}},"created_at":"2026-07-05T01:45:03.927595+00:00","updated_at":"2026-07-05T01:45:03.927595+00:00"}