{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:6J3FMEQSXDJU5TYJ2V4NOWJJKX","short_pith_number":"pith:6J3FMEQS","schema_version":"1.0","canonical_sha256":"f276561212b8d34ecf09d578d7592955ee36d56055ba11cd486708cbb19dfffe","source":{"kind":"arxiv","id":"2212.00196","version":2},"attestation_state":"computed","paper":{"title":"Data-Efficient Finetuning Using Cross-Task Nearest Neighbors","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hamish Ivison, Hannaneh Hajishirzi, Noah A. Smith, Pradeep Dasigi","submitted_at":"2022-12-01T00:53:04Z","abstract_excerpt":"Obtaining labeled data to train a model for a task of interest is often expensive. Prior work shows training models on multitask data augmented with task descriptions (prompts) effectively transfers knowledge to new tasks. Towards efficiently building task-specific models, we assume access to a small number (32-1000) of unlabeled target-task examples and use those to retrieve the most similar labeled examples from a large pool of multitask data augmented with prompts. Compared to the current practice of finetuning models on uniformly sampled prompted multitask data (e.g.: FLAN, T0), our approa"},"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":"2212.00196","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-12-01T00:53:04Z","cross_cats_sorted":[],"title_canon_sha256":"cb7624d13109dc3edffd4d10a2aead797fe2fa7fba021394da7520a8d24f63d7","abstract_canon_sha256":"4e5f9adb527fa69411a723b0e074f9ac151e2405c87f58c0a013a58ce05c2fbc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:13:37.501586Z","signature_b64":"K9D/FrkVQxhi9MtIH/ZWwfZ2QdXLLXbUXpSjhP1nzJ1g9WMvSd0IzuqlPcktqu79cj7xL9y4WGIoJ7KWpYWxDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f276561212b8d34ecf09d578d7592955ee36d56055ba11cd486708cbb19dfffe","last_reissued_at":"2026-07-05T06:13:37.501083Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:13:37.501083Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Data-Efficient Finetuning Using Cross-Task Nearest Neighbors","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Hamish Ivison, Hannaneh Hajishirzi, Noah A. Smith, Pradeep Dasigi","submitted_at":"2022-12-01T00:53:04Z","abstract_excerpt":"Obtaining labeled data to train a model for a task of interest is often expensive. Prior work shows training models on multitask data augmented with task descriptions (prompts) effectively transfers knowledge to new tasks. Towards efficiently building task-specific models, we assume access to a small number (32-1000) of unlabeled target-task examples and use those to retrieve the most similar labeled examples from a large pool of multitask data augmented with prompts. Compared to the current practice of finetuning models on uniformly sampled prompted multitask data (e.g.: FLAN, T0), our approa"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2212.00196","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/2212.00196/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":"2212.00196","created_at":"2026-07-05T06:13:37.501143+00:00"},{"alias_kind":"arxiv_version","alias_value":"2212.00196v2","created_at":"2026-07-05T06:13:37.501143+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2212.00196","created_at":"2026-07-05T06:13:37.501143+00:00"},{"alias_kind":"pith_short_12","alias_value":"6J3FMEQSXDJU","created_at":"2026-07-05T06:13:37.501143+00:00"},{"alias_kind":"pith_short_16","alias_value":"6J3FMEQSXDJU5TYJ","created_at":"2026-07-05T06:13:37.501143+00:00"},{"alias_kind":"pith_short_8","alias_value":"6J3FMEQS","created_at":"2026-07-05T06:13:37.501143+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.22389","citing_title":"Unified Data Selection for LLM Reasoning","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX","json":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX.json","graph_json":"https://pith.science/api/pith-number/6J3FMEQSXDJU5TYJ2V4NOWJJKX/graph.json","events_json":"https://pith.science/api/pith-number/6J3FMEQSXDJU5TYJ2V4NOWJJKX/events.json","paper":"https://pith.science/paper/6J3FMEQS"},"agent_actions":{"view_html":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX","download_json":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX.json","view_paper":"https://pith.science/paper/6J3FMEQS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2212.00196&json=true","fetch_graph":"https://pith.science/api/pith-number/6J3FMEQSXDJU5TYJ2V4NOWJJKX/graph.json","fetch_events":"https://pith.science/api/pith-number/6J3FMEQSXDJU5TYJ2V4NOWJJKX/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX/action/storage_attestation","attest_author":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX/action/author_attestation","sign_citation":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX/action/citation_signature","submit_replication":"https://pith.science/pith/6J3FMEQSXDJU5TYJ2V4NOWJJKX/action/replication_record"}},"created_at":"2026-07-05T06:13:37.501143+00:00","updated_at":"2026-07-05T06:13:37.501143+00:00"}