{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:CXH2XFXPNHE7OOVQNJ4RIITPUQ","short_pith_number":"pith:CXH2XFXP","schema_version":"1.0","canonical_sha256":"15cfab96ef69c9f73ab06a7914226fa40834641724768631d76d6e2f8fd981c3","source":{"kind":"arxiv","id":"2205.10782","version":1},"attestation_state":"computed","paper":{"title":"Instruction Induction: From Few Examples to Natural Language Task Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Omer Levy, Or Honovich, Samuel R. Bowman, Uri Shaham","submitted_at":"2022-05-22T09:22:37Z","abstract_excerpt":"Large language models are able to perform a task by conditioning on a few input-output demonstrations - a paradigm known as in-context learning. We show that language models can explicitly infer an underlying task from a few demonstrations by prompting them to generate a natural language instruction that fits the examples. To explore this ability, we introduce the instruction induction challenge, compile a dataset consisting of 24 tasks, and define a novel evaluation metric based on executing the generated instruction. We discover that, to a large extent, the ability to generate instructions d"},"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":"2205.10782","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-05-22T09:22:37Z","cross_cats_sorted":[],"title_canon_sha256":"2899dd2976dd11dd057ef17c1a83a5dc5acaa2f95547a46b88e2d2d5dca30c25","abstract_canon_sha256":"d9df45e9466060bb9cb62265e637c94483e5b25fa0338958ac276f5751dbea7b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T04:25:23.459789Z","signature_b64":"sasWDvMsE0ucattf58FEuNA4RuUW38rA7raE14CHv6qOk8iV3nQM77K4g7neRtCEQuFrMrO84RAveEkzftbkAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"15cfab96ef69c9f73ab06a7914226fa40834641724768631d76d6e2f8fd981c3","last_reissued_at":"2026-07-05T04:25:23.459301Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T04:25:23.459301Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Instruction Induction: From Few Examples to Natural Language Task Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Omer Levy, Or Honovich, Samuel R. Bowman, Uri Shaham","submitted_at":"2022-05-22T09:22:37Z","abstract_excerpt":"Large language models are able to perform a task by conditioning on a few input-output demonstrations - a paradigm known as in-context learning. We show that language models can explicitly infer an underlying task from a few demonstrations by prompting them to generate a natural language instruction that fits the examples. To explore this ability, we introduce the instruction induction challenge, compile a dataset consisting of 24 tasks, and define a novel evaluation metric based on executing the generated instruction. We discover that, to a large extent, the ability to generate instructions d"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2205.10782","kind":"arxiv","version":1},"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/2205.10782/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":"2205.10782","created_at":"2026-07-05T04:25:23.459369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2205.10782v1","created_at":"2026-07-05T04:25:23.459369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2205.10782","created_at":"2026-07-05T04:25:23.459369+00:00"},{"alias_kind":"pith_short_12","alias_value":"CXH2XFXPNHE7","created_at":"2026-07-05T04:25:23.459369+00:00"},{"alias_kind":"pith_short_16","alias_value":"CXH2XFXPNHE7OOVQ","created_at":"2026-07-05T04:25:23.459369+00:00"},{"alias_kind":"pith_short_8","alias_value":"CXH2XFXP","created_at":"2026-07-05T04:25:23.459369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2211.01910","citing_title":"Large Language Models Are Human-Level Prompt Engineers","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2507.20906","citing_title":"Soft Head Selection for Injecting ICL-Derived Task Embeddings","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2303.09014","citing_title":"ART: Automatic multi-step reasoning and tool-use for large language models","ref_index":153,"is_internal_anchor":false},{"citing_arxiv_id":"2309.08532","citing_title":"EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers","ref_index":122,"is_internal_anchor":false},{"citing_arxiv_id":"2604.04942","citing_title":"TDA-RC: Task-Driven Alignment for Knowledge-Based Reasoning Chains in Large Language Models","ref_index":46,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ","json":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ.json","graph_json":"https://pith.science/api/pith-number/CXH2XFXPNHE7OOVQNJ4RIITPUQ/graph.json","events_json":"https://pith.science/api/pith-number/CXH2XFXPNHE7OOVQNJ4RIITPUQ/events.json","paper":"https://pith.science/paper/CXH2XFXP"},"agent_actions":{"view_html":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ","download_json":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ.json","view_paper":"https://pith.science/paper/CXH2XFXP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2205.10782&json=true","fetch_graph":"https://pith.science/api/pith-number/CXH2XFXPNHE7OOVQNJ4RIITPUQ/graph.json","fetch_events":"https://pith.science/api/pith-number/CXH2XFXPNHE7OOVQNJ4RIITPUQ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ/action/storage_attestation","attest_author":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ/action/author_attestation","sign_citation":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ/action/citation_signature","submit_replication":"https://pith.science/pith/CXH2XFXPNHE7OOVQNJ4RIITPUQ/action/replication_record"}},"created_at":"2026-07-05T04:25:23.459369+00:00","updated_at":"2026-07-05T04:25:23.459369+00:00"}