{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EYCLPWTQQ5PTFO6Z7RSF3LDBVJ","short_pith_number":"pith:EYCLPWTQ","schema_version":"1.0","canonical_sha256":"2604b7da70875f32bbd9fc645dac61aa6046093b419087993cb9d5ced8e883d5","source":{"kind":"arxiv","id":"2603.11679","version":3},"attestation_state":"computed","paper":{"title":"LLMs can construct powerful representations and streamline sample-efficient supervised learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"David Sontag, Ilker Demirel, Lawrence Shi, Zeshan Hussain","submitted_at":"2026-03-12T08:44:06Z","abstract_excerpt":"As real-world datasets become more complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data, such as time-series, free text, and structured records, often requires non-trivial domain expertise. We propose an agentic pipeline to streamline this process. First, an LLM analyzes a small but diverse subset of text-serialized input examples in-context to synthesize a global rubric, which acts as a programmatic specification for extracting and organizing evidence. This rubric is then used to transform naive text-serializations of in"},"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":"2603.11679","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-03-12T08:44:06Z","cross_cats_sorted":[],"title_canon_sha256":"0f866a8f167ce6d21c5a063a403c6d28bcc7da6dac5d841e1e684c89ae587b6b","abstract_canon_sha256":"2e96a02dccad3457b90e6c4e19e2d56c37a73aacb411ccd7947c8b3c133d8e1c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-22T01:03:18.703011Z","signature_b64":"eoyP2A6h/HZFcm3okbOIDCqsEBmpw+Tw0SmgmYpzInSIX12lRld+1AtM5f5Df/X702xRMhGMsS3KLIOGdKxvAw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2604b7da70875f32bbd9fc645dac61aa6046093b419087993cb9d5ced8e883d5","last_reissued_at":"2026-05-22T01:03:18.702144Z","signature_status":"signed_v1","first_computed_at":"2026-05-22T01:03:18.702144Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"LLMs can construct powerful representations and streamline sample-efficient supervised learning","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"David Sontag, Ilker Demirel, Lawrence Shi, Zeshan Hussain","submitted_at":"2026-03-12T08:44:06Z","abstract_excerpt":"As real-world datasets become more complex and heterogeneous, supervised learning is often bottlenecked by input representation design. Modeling multimodal data, such as time-series, free text, and structured records, often requires non-trivial domain expertise. We propose an agentic pipeline to streamline this process. First, an LLM analyzes a small but diverse subset of text-serialized input examples in-context to synthesize a global rubric, which acts as a programmatic specification for extracting and organizing evidence. This rubric is then used to transform naive text-serializations of in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.11679","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/2603.11679/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":"2603.11679","created_at":"2026-05-22T01:03:18.702239+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.11679v3","created_at":"2026-05-22T01:03:18.702239+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.11679","created_at":"2026-05-22T01:03:18.702239+00:00"},{"alias_kind":"pith_short_12","alias_value":"EYCLPWTQQ5PT","created_at":"2026-05-22T01:03:18.702239+00:00"},{"alias_kind":"pith_short_16","alias_value":"EYCLPWTQQ5PTFO6Z","created_at":"2026-05-22T01:03:18.702239+00:00"},{"alias_kind":"pith_short_8","alias_value":"EYCLPWTQ","created_at":"2026-05-22T01:03:18.702239+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.21455","citing_title":"Mitigating Label Bias with Interpretable Rubric Embeddings","ref_index":19,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ","json":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ.json","graph_json":"https://pith.science/api/pith-number/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/graph.json","events_json":"https://pith.science/api/pith-number/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/events.json","paper":"https://pith.science/paper/EYCLPWTQ"},"agent_actions":{"view_html":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ","download_json":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ.json","view_paper":"https://pith.science/paper/EYCLPWTQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.11679&json=true","fetch_graph":"https://pith.science/api/pith-number/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/graph.json","fetch_events":"https://pith.science/api/pith-number/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/action/storage_attestation","attest_author":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/action/author_attestation","sign_citation":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/action/citation_signature","submit_replication":"https://pith.science/pith/EYCLPWTQQ5PTFO6Z7RSF3LDBVJ/action/replication_record"}},"created_at":"2026-05-22T01:03:18.702239+00:00","updated_at":"2026-05-22T01:03:18.702239+00:00"}