{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:75G4KCUCSTLD3KP6J467QXYFXD","short_pith_number":"pith:75G4KCUC","schema_version":"1.0","canonical_sha256":"ff4dc50a8294d63da9fe4f3df85f05b8ed8425906f0a52edb5de1662a51b253a","source":{"kind":"arxiv","id":"2305.00948","version":4},"attestation_state":"computed","paper":{"title":"Large Linguistic Models: Investigating LLMs' metalinguistic abilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ga\\v{s}per Begu\\v{s}, Maksymilian D\\k{a}bkowski, Ryan Rhodes","submitted_at":"2023-05-01T17:09:33Z","abstract_excerpt":"The performance of large language models (LLMs) has recently improved to the point where models can perform well on many language tasks. We show here that--for the first time--the models can also generate valid metalinguistic analyses of language data. We outline a research program where the behavioral interpretability of LLMs on these tasks is tested via prompting. LLMs are trained primarily on text--as such, evaluating their metalinguistic abilities improves our understanding of their general capabilities and sheds new light on theoretical models in linguistics. We show that OpenAI's (2024) "},"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":"2305.00948","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-05-01T17:09:33Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"72de90d04d03f44ec8cdb1e9103aeb88184b842860bc6e10253e263156539bac","abstract_canon_sha256":"c728f93a4c3c3afb19f3c8c0abad253f2ca267cbe4487df8d93f4758a2604f0d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:35:27.634877Z","signature_b64":"DW9ZmntqNVmdj/82uy7M+oMLvBNDHgKmlrjIFXiuwCvg1OT7zJTZm19qCS5hHfUDtmzNAGICz46BI7nOhJgYCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ff4dc50a8294d63da9fe4f3df85f05b8ed8425906f0a52edb5de1662a51b253a","last_reissued_at":"2026-07-05T11:35:27.634360Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:35:27.634360Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Large Linguistic Models: Investigating LLMs' metalinguistic abilities","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Ga\\v{s}per Begu\\v{s}, Maksymilian D\\k{a}bkowski, Ryan Rhodes","submitted_at":"2023-05-01T17:09:33Z","abstract_excerpt":"The performance of large language models (LLMs) has recently improved to the point where models can perform well on many language tasks. We show here that--for the first time--the models can also generate valid metalinguistic analyses of language data. We outline a research program where the behavioral interpretability of LLMs on these tasks is tested via prompting. LLMs are trained primarily on text--as such, evaluating their metalinguistic abilities improves our understanding of their general capabilities and sheds new light on theoretical models in linguistics. We show that OpenAI's (2024) "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.00948","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":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2305.00948/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":"2305.00948","created_at":"2026-07-05T11:35:27.634424+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.00948v4","created_at":"2026-07-05T11:35:27.634424+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.00948","created_at":"2026-07-05T11:35:27.634424+00:00"},{"alias_kind":"pith_short_12","alias_value":"75G4KCUCSTLD","created_at":"2026-07-05T11:35:27.634424+00:00"},{"alias_kind":"pith_short_16","alias_value":"75G4KCUCSTLD3KP6","created_at":"2026-07-05T11:35:27.634424+00:00"},{"alias_kind":"pith_short_8","alias_value":"75G4KCUC","created_at":"2026-07-05T11:35:27.634424+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.01045","citing_title":"Child-directed speech facilitates production, not comprehension, in BabyLMs","ref_index":61,"is_internal_anchor":false},{"citing_arxiv_id":"2412.14171","citing_title":"Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2505.20279","citing_title":"VLM-3R: Vision-Language Models Augmented with Instruction-Aligned 3D Reconstruction","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11206","citing_title":"Instructions Shape Production of Language, not Processing","ref_index":259,"is_internal_anchor":false},{"citing_arxiv_id":"2605.11206","citing_title":"Instructions Shape Production of Language, not Processing","ref_index":259,"is_internal_anchor":false},{"citing_arxiv_id":"2604.09535","citing_title":"EgoTL: Egocentric Think-Aloud Chains for Long-Horizon Tasks","ref_index":6,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD","json":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD.json","graph_json":"https://pith.science/api/pith-number/75G4KCUCSTLD3KP6J467QXYFXD/graph.json","events_json":"https://pith.science/api/pith-number/75G4KCUCSTLD3KP6J467QXYFXD/events.json","paper":"https://pith.science/paper/75G4KCUC"},"agent_actions":{"view_html":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD","download_json":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD.json","view_paper":"https://pith.science/paper/75G4KCUC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.00948&json=true","fetch_graph":"https://pith.science/api/pith-number/75G4KCUCSTLD3KP6J467QXYFXD/graph.json","fetch_events":"https://pith.science/api/pith-number/75G4KCUCSTLD3KP6J467QXYFXD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD/action/storage_attestation","attest_author":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD/action/author_attestation","sign_citation":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD/action/citation_signature","submit_replication":"https://pith.science/pith/75G4KCUCSTLD3KP6J467QXYFXD/action/replication_record"}},"created_at":"2026-07-05T11:35:27.634424+00:00","updated_at":"2026-07-05T11:35:27.634424+00:00"}