{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:D3KLDYRDTKIYKSQT6KVQ2QD64R","short_pith_number":"pith:D3KLDYRD","schema_version":"1.0","canonical_sha256":"1ed4b1e2239a91854a13f2ab0d407ee457c42ebcc5599e4dfc0c71400d95f3d6","source":{"kind":"arxiv","id":"2103.11955","version":3},"attestation_state":"computed","paper":{"title":"Improving and Simplifying Pattern Exploiting Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Colin Raffel, Derek Tam, Mohit Bansal, Rakesh R Menon, Shashank Srivastava","submitted_at":"2021-03-22T15:52:45Z","abstract_excerpt":"Recently, pre-trained language models (LMs) have achieved strong performance when fine-tuned on difficult benchmarks like SuperGLUE. However, performance can suffer when there are very few labeled examples available for fine-tuning. Pattern Exploiting Training (PET) is a recent approach that leverages patterns for few-shot learning. However, PET uses task-specific unlabeled data. In this paper, we focus on few-shot learning without any unlabeled data and introduce ADAPET, which modifies PET's objective to provide denser supervision during fine-tuning. As a result, ADAPET outperforms PET on Sup"},"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":"2103.11955","kind":"arxiv","version":3},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-03-22T15:52:45Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"4024c9ec7818216cd15d1660a58a4f7a8d4236aee0b51aedb6ff830f0d83c17a","abstract_canon_sha256":"4fcff6661eab29871e1c8f364a358cd923fa19e8400ba3ecb9fe5258e0c7c9a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:17:49.131113Z","signature_b64":"7enRmaRztrU3FZB5bDpI0DbUhJnfTlAPOC6tXdm9wxnFYt4VCiCoPBssNFxRQdgBVW82V/+L7VBqnCru41mwDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"1ed4b1e2239a91854a13f2ab0d407ee457c42ebcc5599e4dfc0c71400d95f3d6","last_reissued_at":"2026-07-05T03:17:49.130610Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:17:49.130610Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Improving and Simplifying Pattern Exploiting Training","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Colin Raffel, Derek Tam, Mohit Bansal, Rakesh R Menon, Shashank Srivastava","submitted_at":"2021-03-22T15:52:45Z","abstract_excerpt":"Recently, pre-trained language models (LMs) have achieved strong performance when fine-tuned on difficult benchmarks like SuperGLUE. However, performance can suffer when there are very few labeled examples available for fine-tuning. Pattern Exploiting Training (PET) is a recent approach that leverages patterns for few-shot learning. However, PET uses task-specific unlabeled data. In this paper, we focus on few-shot learning without any unlabeled data and introduce ADAPET, which modifies PET's objective to provide denser supervision during fine-tuning. As a result, ADAPET outperforms PET on Sup"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2103.11955","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/2103.11955/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":"2103.11955","created_at":"2026-07-05T03:17:49.130674+00:00"},{"alias_kind":"arxiv_version","alias_value":"2103.11955v3","created_at":"2026-07-05T03:17:49.130674+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2103.11955","created_at":"2026-07-05T03:17:49.130674+00:00"},{"alias_kind":"pith_short_12","alias_value":"D3KLDYRDTKIY","created_at":"2026-07-05T03:17:49.130674+00:00"},{"alias_kind":"pith_short_16","alias_value":"D3KLDYRDTKIYKSQT","created_at":"2026-07-05T03:17:49.130674+00:00"},{"alias_kind":"pith_short_8","alias_value":"D3KLDYRD","created_at":"2026-07-05T03:17:49.130674+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2104.08773","citing_title":"Cross-Task Generalization via Natural Language Crowdsourcing Instructions","ref_index":22,"is_internal_anchor":false},{"citing_arxiv_id":"2208.03299","citing_title":"Atlas: Few-shot Learning with Retrieval Augmented Language Models","ref_index":250,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R","json":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R.json","graph_json":"https://pith.science/api/pith-number/D3KLDYRDTKIYKSQT6KVQ2QD64R/graph.json","events_json":"https://pith.science/api/pith-number/D3KLDYRDTKIYKSQT6KVQ2QD64R/events.json","paper":"https://pith.science/paper/D3KLDYRD"},"agent_actions":{"view_html":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R","download_json":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R.json","view_paper":"https://pith.science/paper/D3KLDYRD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2103.11955&json=true","fetch_graph":"https://pith.science/api/pith-number/D3KLDYRDTKIYKSQT6KVQ2QD64R/graph.json","fetch_events":"https://pith.science/api/pith-number/D3KLDYRDTKIYKSQT6KVQ2QD64R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R/action/storage_attestation","attest_author":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R/action/author_attestation","sign_citation":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R/action/citation_signature","submit_replication":"https://pith.science/pith/D3KLDYRDTKIYKSQT6KVQ2QD64R/action/replication_record"}},"created_at":"2026-07-05T03:17:49.130674+00:00","updated_at":"2026-07-05T03:17:49.130674+00:00"}