{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:RSHOSWZRQ66BA2YL74VPTJZCI5","short_pith_number":"pith:RSHOSWZR","schema_version":"1.0","canonical_sha256":"8c8ee95b3187bc106b0bff2af9a722475da07c373ce8e083144a73389db98b86","source":{"kind":"arxiv","id":"2501.06346","version":2},"attestation_state":"computed","paper":{"title":"Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Mueller, Christian Bartelt, Chris Wendler, Jannik Brinkmann","submitted_at":"2025-01-10T21:18:21Z","abstract_excerpt":"Human bilinguals often use similar brain regions to process multiple languages, depending on when they learned their second language and their proficiency. In large language models (LLMs), how are multiple languages learned and encoded? In this work, we explore the extent to which LLMs share representations of morphsyntactic concepts such as grammatical number, gender, and tense across languages. We train sparse autoencoders on Llama-3-8B and Aya-23-8B, and demonstrate that abstract grammatical concepts are often encoded in feature directions shared across many languages. We use causal interve"},"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":"2501.06346","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-01-10T21:18:21Z","cross_cats_sorted":[],"title_canon_sha256":"b6c6c96697910f25d8d7f0bd9de12d92885ad652dd383d3cf4cec9556b0a2f5d","abstract_canon_sha256":"cf2847901990cba24cec56f3cff50f1e1bbe6885e3a401177a493b2973780810"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:08:15.106097Z","signature_b64":"9QhWwQVx4Sjw1vvjOOqFa4q5HVXF/tEaizQPQ0ZpUCCtWrpxX2UgjdbQNBB79TSl/VcFoN0aYykYU7m0JZVUDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8c8ee95b3187bc106b0bff2af9a722475da07c373ce8e083144a73389db98b86","last_reissued_at":"2026-07-05T11:08:15.105537Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:08:15.105537Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Large Language Models Share Representations of Latent Grammatical Concepts Across Typologically Diverse Languages","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Aaron Mueller, Christian Bartelt, Chris Wendler, Jannik Brinkmann","submitted_at":"2025-01-10T21:18:21Z","abstract_excerpt":"Human bilinguals often use similar brain regions to process multiple languages, depending on when they learned their second language and their proficiency. In large language models (LLMs), how are multiple languages learned and encoded? In this work, we explore the extent to which LLMs share representations of morphsyntactic concepts such as grammatical number, gender, and tense across languages. We train sparse autoencoders on Llama-3-8B and Aya-23-8B, and demonstrate that abstract grammatical concepts are often encoded in feature directions shared across many languages. We use causal interve"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.06346","kind":"arxiv","version":2},"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/2501.06346/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":"2501.06346","created_at":"2026-07-05T11:08:15.105597+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.06346v2","created_at":"2026-07-05T11:08:15.105597+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.06346","created_at":"2026-07-05T11:08:15.105597+00:00"},{"alias_kind":"pith_short_12","alias_value":"RSHOSWZRQ66B","created_at":"2026-07-05T11:08:15.105597+00:00"},{"alias_kind":"pith_short_16","alias_value":"RSHOSWZRQ66BA2YL","created_at":"2026-07-05T11:08:15.105597+00:00"},{"alias_kind":"pith_short_8","alias_value":"RSHOSWZR","created_at":"2026-07-05T11:08:15.105597+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2512.03676","citing_title":"Different types of syntactic agreement recruit the same units within large language models","ref_index":7,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5","json":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5.json","graph_json":"https://pith.science/api/pith-number/RSHOSWZRQ66BA2YL74VPTJZCI5/graph.json","events_json":"https://pith.science/api/pith-number/RSHOSWZRQ66BA2YL74VPTJZCI5/events.json","paper":"https://pith.science/paper/RSHOSWZR"},"agent_actions":{"view_html":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5","download_json":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5.json","view_paper":"https://pith.science/paper/RSHOSWZR","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.06346&json=true","fetch_graph":"https://pith.science/api/pith-number/RSHOSWZRQ66BA2YL74VPTJZCI5/graph.json","fetch_events":"https://pith.science/api/pith-number/RSHOSWZRQ66BA2YL74VPTJZCI5/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5/action/storage_attestation","attest_author":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5/action/author_attestation","sign_citation":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5/action/citation_signature","submit_replication":"https://pith.science/pith/RSHOSWZRQ66BA2YL74VPTJZCI5/action/replication_record"}},"created_at":"2026-07-05T11:08:15.105597+00:00","updated_at":"2026-07-05T11:08:15.105597+00:00"}