{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:T4KJ7WHQK2DHRHGGFZ6QE75SSS","short_pith_number":"pith:T4KJ7WHQ","schema_version":"1.0","canonical_sha256":"9f149fd8f05686789cc62e7d027fb2948278a4c364c2e229e943aa183706648d","source":{"kind":"arxiv","id":"2606.30815","version":1},"attestation_state":"computed","paper":{"title":"When transformers learn \"impossible\" languages, what do they learn?","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Coleman Haley, Ram Janarthan, Sharon Goldwater","submitted_at":"2026-06-29T18:42:03Z","abstract_excerpt":"Recent work suggests that transformer language models show a bias towards human languages over unnatural (\"impossible\") languages argued to be unacquirable by humans. However, this literature has largely based these claims on differences in sample efficiency and test-set perplexity, rather than on direct evaluations of the linguistic capacities that could plausibly explain non-attestation in human languages. We evaluate two theoretically motivated linking hypotheses: impossibility arising from deficiencies in grammatical sensitivity or generative production. Using GPT-2 style models trained on"},"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":"2606.30815","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-29T18:42:03Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"177ef10be9b78d9cf41e15eca33d26242935b93f6575149b82c500d3f121e869","abstract_canon_sha256":"0ad8715aa4f66a7d1b8ff83eb3918cbb3fbc82662dce7b9b0485809e57fca058"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T00:17:18.349642Z","signature_b64":"1pUSPVQ/AHGcOslcg9uUWxuDEkVFL06vD1rxw27VQBgMHdWGxgIBAquZEnbn0PaWAmjNoVwv9thf5ltRdDruCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9f149fd8f05686789cc62e7d027fb2948278a4c364c2e229e943aa183706648d","last_reissued_at":"2026-07-01T00:17:18.349223Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T00:17:18.349223Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"When transformers learn \"impossible\" languages, what do they learn?","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Coleman Haley, Ram Janarthan, Sharon Goldwater","submitted_at":"2026-06-29T18:42:03Z","abstract_excerpt":"Recent work suggests that transformer language models show a bias towards human languages over unnatural (\"impossible\") languages argued to be unacquirable by humans. However, this literature has largely based these claims on differences in sample efficiency and test-set perplexity, rather than on direct evaluations of the linguistic capacities that could plausibly explain non-attestation in human languages. We evaluate two theoretically motivated linking hypotheses: impossibility arising from deficiencies in grammatical sensitivity or generative production. Using GPT-2 style models trained on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.30815","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/2606.30815/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":"2606.30815","created_at":"2026-07-01T00:17:18.349280+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.30815v1","created_at":"2026-07-01T00:17:18.349280+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.30815","created_at":"2026-07-01T00:17:18.349280+00:00"},{"alias_kind":"pith_short_12","alias_value":"T4KJ7WHQK2DH","created_at":"2026-07-01T00:17:18.349280+00:00"},{"alias_kind":"pith_short_16","alias_value":"T4KJ7WHQK2DHRHGG","created_at":"2026-07-01T00:17:18.349280+00:00"},{"alias_kind":"pith_short_8","alias_value":"T4KJ7WHQ","created_at":"2026-07-01T00:17:18.349280+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS","json":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS.json","graph_json":"https://pith.science/api/pith-number/T4KJ7WHQK2DHRHGGFZ6QE75SSS/graph.json","events_json":"https://pith.science/api/pith-number/T4KJ7WHQK2DHRHGGFZ6QE75SSS/events.json","paper":"https://pith.science/paper/T4KJ7WHQ"},"agent_actions":{"view_html":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS","download_json":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS.json","view_paper":"https://pith.science/paper/T4KJ7WHQ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.30815&json=true","fetch_graph":"https://pith.science/api/pith-number/T4KJ7WHQK2DHRHGGFZ6QE75SSS/graph.json","fetch_events":"https://pith.science/api/pith-number/T4KJ7WHQK2DHRHGGFZ6QE75SSS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS/action/storage_attestation","attest_author":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS/action/author_attestation","sign_citation":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS/action/citation_signature","submit_replication":"https://pith.science/pith/T4KJ7WHQK2DHRHGGFZ6QE75SSS/action/replication_record"}},"created_at":"2026-07-01T00:17:18.349280+00:00","updated_at":"2026-07-01T00:17:18.349280+00:00"}