{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:QDJYZSFM7KDCPV6MJQGC3PNYJD","short_pith_number":"pith:QDJYZSFM","schema_version":"1.0","canonical_sha256":"80d38cc8acfa8627d7cc4c0c2dbdb848d1fadee5559b8ab4a533080c8e0b8988","source":{"kind":"arxiv","id":"2104.08773","version":4},"attestation_state":"computed","paper":{"title":"Cross-Task Generalization via Natural Language Crowdsourcing Instructions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chitta Baral, Daniel Khashabi, Hannaneh Hajishirzi, Swaroop Mishra","submitted_at":"2021-04-18T08:44:56Z","abstract_excerpt":"Humans (e.g., crowdworkers) have a remarkable ability in solving different tasks, by simply reading textual instructions that define them and looking at a few examples. Despite the success of the conventional supervised learning on individual datasets, such models often struggle with generalization across tasks (e.g., a question-answering system cannot solve classification tasks). A long-standing challenge in AI is to build a model that learns a new task by understanding the human-readable instructions that define it. To study this, we introduce NATURAL INSTRUCTIONS, a dataset of 61 distinct 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":"2104.08773","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2021-04-18T08:44:56Z","cross_cats_sorted":["cs.AI","cs.CV","cs.LG"],"title_canon_sha256":"95a7184b85e9c0acf0839b00dfa8bd0e19dfbc744a13acf41a6e6745227993fe","abstract_canon_sha256":"c2569054dca373edb892d0e88e982fc12341cd8b9da3796b2b998721fa029d8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T01:52:20.055975Z","signature_b64":"vI+Rr/raG9kXT6Es0eprj0EBj2FSlK9TZAQ64sCERHODnblS1s2eiRWulHpAOstvF2z9XAcLSQG0syEjptRFCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"80d38cc8acfa8627d7cc4c0c2dbdb848d1fadee5559b8ab4a533080c8e0b8988","last_reissued_at":"2026-05-18T01:52:20.055403Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T01:52:20.055403Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Cross-Task Generalization via Natural Language Crowdsourcing Instructions","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CV","cs.LG"],"primary_cat":"cs.CL","authors_text":"Chitta Baral, Daniel Khashabi, Hannaneh Hajishirzi, Swaroop Mishra","submitted_at":"2021-04-18T08:44:56Z","abstract_excerpt":"Humans (e.g., crowdworkers) have a remarkable ability in solving different tasks, by simply reading textual instructions that define them and looking at a few examples. Despite the success of the conventional supervised learning on individual datasets, such models often struggle with generalization across tasks (e.g., a question-answering system cannot solve classification tasks). A long-standing challenge in AI is to build a model that learns a new task by understanding the human-readable instructions that define it. To study this, we introduce NATURAL INSTRUCTIONS, a dataset of 61 distinct t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2104.08773","kind":"arxiv","version":4},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"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":"2104.08773","created_at":"2026-05-18T01:52:20.055483+00:00"},{"alias_kind":"arxiv_version","alias_value":"2104.08773v4","created_at":"2026-05-18T01:52:20.055483+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2104.08773","created_at":"2026-05-18T01:52:20.055483+00:00"},{"alias_kind":"pith_short_12","alias_value":"QDJYZSFM7KDC","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_16","alias_value":"QDJYZSFM7KDCPV6M","created_at":"2026-05-18T12:33:33.725879+00:00"},{"alias_kind":"pith_short_8","alias_value":"QDJYZSFM","created_at":"2026-05-18T12:33:33.725879+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":16,"internal_anchor_count":16,"sample":[{"citing_arxiv_id":"2402.13228","citing_title":"Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive","ref_index":37,"is_internal_anchor":true},{"citing_arxiv_id":"2305.11206","citing_title":"LIMA: Less Is More for Alignment","ref_index":44,"is_internal_anchor":true},{"citing_arxiv_id":"2512.11013","citing_title":"PIAST: Rapid Prompting with In-context Augmentation for Scarce Training data","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2303.09014","citing_title":"ART: Automatic multi-step reasoning and tool-use for large language models","ref_index":159,"is_internal_anchor":true},{"citing_arxiv_id":"2405.14782","citing_title":"Lessons from the Trenches on Reproducible Evaluation of Language Models","ref_index":69,"is_internal_anchor":true},{"citing_arxiv_id":"2308.03958","citing_title":"Simple synthetic data reduces sycophancy in large language models","ref_index":23,"is_internal_anchor":true},{"citing_arxiv_id":"2305.16355","citing_title":"PandaGPT: One Model To Instruction-Follow Them All","ref_index":13,"is_internal_anchor":true},{"citing_arxiv_id":"2202.12837","citing_title":"Rethinking the Role of Demonstrations: What Makes In-Context Learning Work?","ref_index":228,"is_internal_anchor":true},{"citing_arxiv_id":"2110.08207","citing_title":"Multitask Prompted Training Enables Zero-Shot Task Generalization","ref_index":38,"is_internal_anchor":true},{"citing_arxiv_id":"2305.06355","citing_title":"VideoChat: Chat-Centric Video Understanding","ref_index":28,"is_internal_anchor":true},{"citing_arxiv_id":"2502.02737","citing_title":"SmolLM2: When Smol Goes Big -- Data-Centric Training of a Small Language Model","ref_index":203,"is_internal_anchor":true},{"citing_arxiv_id":"2401.10020","citing_title":"Self-Rewarding Language Models","ref_index":35,"is_internal_anchor":true},{"citing_arxiv_id":"2406.01574","citing_title":"MMLU-Pro: A More Robust and Challenging Multi-Task Language Understanding Benchmark","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2402.06196","citing_title":"Large Language Models: A Survey","ref_index":134,"is_internal_anchor":true},{"citing_arxiv_id":"2205.01068","citing_title":"OPT: Open Pre-trained Transformer Language Models","ref_index":205,"is_internal_anchor":true},{"citing_arxiv_id":"2403.07974","citing_title":"LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code","ref_index":184,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD","json":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD.json","graph_json":"https://pith.science/api/pith-number/QDJYZSFM7KDCPV6MJQGC3PNYJD/graph.json","events_json":"https://pith.science/api/pith-number/QDJYZSFM7KDCPV6MJQGC3PNYJD/events.json","paper":"https://pith.science/paper/QDJYZSFM"},"agent_actions":{"view_html":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD","download_json":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD.json","view_paper":"https://pith.science/paper/QDJYZSFM","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2104.08773&json=true","fetch_graph":"https://pith.science/api/pith-number/QDJYZSFM7KDCPV6MJQGC3PNYJD/graph.json","fetch_events":"https://pith.science/api/pith-number/QDJYZSFM7KDCPV6MJQGC3PNYJD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD/action/storage_attestation","attest_author":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD/action/author_attestation","sign_citation":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD/action/citation_signature","submit_replication":"https://pith.science/pith/QDJYZSFM7KDCPV6MJQGC3PNYJD/action/replication_record"}},"created_at":"2026-05-18T01:52:20.055483+00:00","updated_at":"2026-05-18T01:52:20.055483+00:00"}