{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:A2JMMUMZ6I5YFBRLKLRSGCMGLO","short_pith_number":"pith:A2JMMUMZ","schema_version":"1.0","canonical_sha256":"0692c65199f23b82862b52e32309865b94ea413cce67f6916b1b9a71a94da863","source":{"kind":"arxiv","id":"2108.11696","version":2},"attestation_state":"computed","paper":{"title":"Rethinking Why Intermediate-Task Fine-Tuning Works","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chi-Jen Lu, Ting-Yun Chang","submitted_at":"2021-08-26T10:34:37Z","abstract_excerpt":"Supplementary Training on Intermediate Labeled-data Tasks (STILTs) is a widely applied technique, which first fine-tunes the pretrained language models on an intermediate task before on the target task of interest. While STILTs is able to further improve the performance of pretrained language models, it is still unclear why and when it works. Previous research shows that those intermediate tasks involving complex inference, such as commonsense reasoning, work especially well for RoBERTa. In this paper, we discover that the improvement from an intermediate task could be orthogonal to it contain"},"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":"2108.11696","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-08-26T10:34:37Z","cross_cats_sorted":[],"title_canon_sha256":"2ea638a6e6a04f0ead0f6ba581187428bcedadb3bffd438124b98d4e9946c30c","abstract_canon_sha256":"00746eafe145f88dc0170640959a54b2a60299f4181206504315368c63afc2ba"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T03:10:41.793420Z","signature_b64":"k5Q+Y+o4oVZs/fG/SuZTk23YA1rUJ22fJM3f99ULtQ1wozHgbJPbdKcnGjsRoQSnmNhWrJO2F6LBl2LrEd5XCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0692c65199f23b82862b52e32309865b94ea413cce67f6916b1b9a71a94da863","last_reissued_at":"2026-07-05T03:10:41.793072Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T03:10:41.793072Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking Why Intermediate-Task Fine-Tuning Works","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Chi-Jen Lu, Ting-Yun Chang","submitted_at":"2021-08-26T10:34:37Z","abstract_excerpt":"Supplementary Training on Intermediate Labeled-data Tasks (STILTs) is a widely applied technique, which first fine-tunes the pretrained language models on an intermediate task before on the target task of interest. While STILTs is able to further improve the performance of pretrained language models, it is still unclear why and when it works. Previous research shows that those intermediate tasks involving complex inference, such as commonsense reasoning, work especially well for RoBERTa. In this paper, we discover that the improvement from an intermediate task could be orthogonal to it contain"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2108.11696","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/2108.11696/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":"2108.11696","created_at":"2026-07-05T03:10:41.793120+00:00"},{"alias_kind":"arxiv_version","alias_value":"2108.11696v2","created_at":"2026-07-05T03:10:41.793120+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2108.11696","created_at":"2026-07-05T03:10:41.793120+00:00"},{"alias_kind":"pith_short_12","alias_value":"A2JMMUMZ6I5Y","created_at":"2026-07-05T03:10:41.793120+00:00"},{"alias_kind":"pith_short_16","alias_value":"A2JMMUMZ6I5YFBRL","created_at":"2026-07-05T03:10:41.793120+00:00"},{"alias_kind":"pith_short_8","alias_value":"A2JMMUMZ","created_at":"2026-07-05T03:10:41.793120+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2504.21685","citing_title":"Enhancing Health Mention Classification Performance: A Study on Advancements in Parameter Efficient Tuning","ref_index":2020,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO","json":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO.json","graph_json":"https://pith.science/api/pith-number/A2JMMUMZ6I5YFBRLKLRSGCMGLO/graph.json","events_json":"https://pith.science/api/pith-number/A2JMMUMZ6I5YFBRLKLRSGCMGLO/events.json","paper":"https://pith.science/paper/A2JMMUMZ"},"agent_actions":{"view_html":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO","download_json":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO.json","view_paper":"https://pith.science/paper/A2JMMUMZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2108.11696&json=true","fetch_graph":"https://pith.science/api/pith-number/A2JMMUMZ6I5YFBRLKLRSGCMGLO/graph.json","fetch_events":"https://pith.science/api/pith-number/A2JMMUMZ6I5YFBRLKLRSGCMGLO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO/action/storage_attestation","attest_author":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO/action/author_attestation","sign_citation":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO/action/citation_signature","submit_replication":"https://pith.science/pith/A2JMMUMZ6I5YFBRLKLRSGCMGLO/action/replication_record"}},"created_at":"2026-07-05T03:10:41.793120+00:00","updated_at":"2026-07-05T03:10:41.793120+00:00"}