{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:VTIGCJ5XFTRYWJS6SL2RVO3EFW","short_pith_number":"pith:VTIGCJ5X","schema_version":"1.0","canonical_sha256":"acd06127b72ce38b265e92f51abb642d840f64fe871f672b13ef6e4d664af02e","source":{"kind":"arxiv","id":"2101.06829","version":2},"attestation_state":"computed","paper":{"title":"Joint Energy-based Model Training for Better Calibrated Natural Language Understanding Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bryan McCann, Caiming Xiong, Ehsan Hosseini-Asl, Tianxing He","submitted_at":"2021-01-18T01:41:31Z","abstract_excerpt":"In this work, we explore joint energy-based model (EBM) training during the finetuning of pretrained text encoders (e.g., Roberta) for natural language understanding (NLU) tasks. Our experiments show that EBM training can help the model reach a better calibration that is competitive to strong baselines, with little or no loss in accuracy. We discuss three variants of energy functions (namely scalar, hidden, and sharp-hidden) that can be defined on top of a text encoder, and compare them in experiments. Due to the discreteness of text data, we adopt noise contrastive estimation (NCE) to train t"},"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":"2101.06829","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-01-18T01:41:31Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"b31a7628c1dc3e69bb757f5cfa60104d24396650e17da704fc77bdf5a285512f","abstract_canon_sha256":"da01361134330dd16e4a05e0e0d791fae7f39f5415f0b17d8f6b703290abdeed"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:16:38.218230Z","signature_b64":"8Q5AtlLGqWfod4TNQ+7Jl+NLyy+3TqOI6TjhRnUm+Ue4s0xl2+SbhVkiyfG7/o1nDkFxBe7bW/Yfu2qQ3jpoAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"acd06127b72ce38b265e92f51abb642d840f64fe871f672b13ef6e4d664af02e","last_reissued_at":"2026-07-05T02:16:38.217843Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:16:38.217843Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Joint Energy-based Model Training for Better Calibrated Natural Language Understanding Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Bryan McCann, Caiming Xiong, Ehsan Hosseini-Asl, Tianxing He","submitted_at":"2021-01-18T01:41:31Z","abstract_excerpt":"In this work, we explore joint energy-based model (EBM) training during the finetuning of pretrained text encoders (e.g., Roberta) for natural language understanding (NLU) tasks. Our experiments show that EBM training can help the model reach a better calibration that is competitive to strong baselines, with little or no loss in accuracy. We discuss three variants of energy functions (namely scalar, hidden, and sharp-hidden) that can be defined on top of a text encoder, and compare them in experiments. Due to the discreteness of text data, we adopt noise contrastive estimation (NCE) to train t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2101.06829","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/2101.06829/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":"2101.06829","created_at":"2026-07-05T02:16:38.217896+00:00"},{"alias_kind":"arxiv_version","alias_value":"2101.06829v2","created_at":"2026-07-05T02:16:38.217896+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2101.06829","created_at":"2026-07-05T02:16:38.217896+00:00"},{"alias_kind":"pith_short_12","alias_value":"VTIGCJ5XFTRY","created_at":"2026-07-05T02:16:38.217896+00:00"},{"alias_kind":"pith_short_16","alias_value":"VTIGCJ5XFTRYWJS6","created_at":"2026-07-05T02:16:38.217896+00:00"},{"alias_kind":"pith_short_8","alias_value":"VTIGCJ5X","created_at":"2026-07-05T02:16:38.217896+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2501.05495","citing_title":"LSEBMCL: A Latent Space Energy-Based Model for Continual Learning","ref_index":22,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW","json":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW.json","graph_json":"https://pith.science/api/pith-number/VTIGCJ5XFTRYWJS6SL2RVO3EFW/graph.json","events_json":"https://pith.science/api/pith-number/VTIGCJ5XFTRYWJS6SL2RVO3EFW/events.json","paper":"https://pith.science/paper/VTIGCJ5X"},"agent_actions":{"view_html":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW","download_json":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW.json","view_paper":"https://pith.science/paper/VTIGCJ5X","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2101.06829&json=true","fetch_graph":"https://pith.science/api/pith-number/VTIGCJ5XFTRYWJS6SL2RVO3EFW/graph.json","fetch_events":"https://pith.science/api/pith-number/VTIGCJ5XFTRYWJS6SL2RVO3EFW/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW/action/storage_attestation","attest_author":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW/action/author_attestation","sign_citation":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW/action/citation_signature","submit_replication":"https://pith.science/pith/VTIGCJ5XFTRYWJS6SL2RVO3EFW/action/replication_record"}},"created_at":"2026-07-05T02:16:38.217896+00:00","updated_at":"2026-07-05T02:16:38.217896+00:00"}