{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:TNPWS2RGMTTJJB23R2XPM63EAA","short_pith_number":"pith:TNPWS2RG","schema_version":"1.0","canonical_sha256":"9b5f696a2664e694875b8eaef67b64000c72193a6c4f87a2e84d92c2a4af1320","source":{"kind":"arxiv","id":"2402.16438","version":2},"attestation_state":"computed","paper":{"title":"Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongdong Zhang, Furu Wei, Haoyang Huang, Ji-Rong Wen, Tianyi Tang, Wenyang Luo, Xiaolei Wang, Xin Zhao","submitted_at":"2024-02-26T09:36:05Z","abstract_excerpt":"Large language models (LLMs) demonstrate remarkable multilingual capabilities without being pre-trained on specially curated multilingual parallel corpora. It remains a challenging problem to explain the underlying mechanisms by which LLMs process multilingual texts. In this paper, we delve into the composition of Transformer architectures in LLMs to pinpoint language-specific regions. Specially, we propose a novel detection method, language activation probability entropy (LAPE), to identify language-specific neurons within LLMs. Based on LAPE, we conduct comprehensive experiments on several r"},"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":"2402.16438","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-26T09:36:05Z","cross_cats_sorted":[],"title_canon_sha256":"b4b88ce8a153156d6258b963db829ed1f84ec88ecd2bd79113dbbb1d0bc7dd55","abstract_canon_sha256":"6a248534322b66045ff69d4f75aed8332a23a01de5838f6bf064096232125d84"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:28:06.998793Z","signature_b64":"tMO+wibCOe/riijNJOj+eSyxlZ8ynaV6PXWD4EDEROzMUh2/ubsDKbv3d44I9dh+VA8O2tjYKgbM0k2DfgGsBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9b5f696a2664e694875b8eaef67b64000c72193a6c4f87a2e84d92c2a4af1320","last_reissued_at":"2026-07-05T08:28:06.998277Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:28:06.998277Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Language-Specific Neurons: The Key to Multilingual Capabilities in Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Dongdong Zhang, Furu Wei, Haoyang Huang, Ji-Rong Wen, Tianyi Tang, Wenyang Luo, Xiaolei Wang, Xin Zhao","submitted_at":"2024-02-26T09:36:05Z","abstract_excerpt":"Large language models (LLMs) demonstrate remarkable multilingual capabilities without being pre-trained on specially curated multilingual parallel corpora. It remains a challenging problem to explain the underlying mechanisms by which LLMs process multilingual texts. In this paper, we delve into the composition of Transformer architectures in LLMs to pinpoint language-specific regions. Specially, we propose a novel detection method, language activation probability entropy (LAPE), to identify language-specific neurons within LLMs. Based on LAPE, we conduct comprehensive experiments on several r"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.16438","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/2402.16438/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":"2402.16438","created_at":"2026-07-05T08:28:06.998346+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.16438v2","created_at":"2026-07-05T08:28:06.998346+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.16438","created_at":"2026-07-05T08:28:06.998346+00:00"},{"alias_kind":"pith_short_12","alias_value":"TNPWS2RGMTTJ","created_at":"2026-07-05T08:28:06.998346+00:00"},{"alias_kind":"pith_short_16","alias_value":"TNPWS2RGMTTJJB23","created_at":"2026-07-05T08:28:06.998346+00:00"},{"alias_kind":"pith_short_8","alias_value":"TNPWS2RG","created_at":"2026-07-05T08:28:06.998346+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2412.12686","citing_title":"Exploring Cross-lingual Latent Transplantation: Mutual Opportunities and Open Challenges","ref_index":45,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08881","citing_title":"Targeted Interpretable Safety Neuron Enhancement for Multilingual Vision-Language Large Models","ref_index":32,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA","json":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA.json","graph_json":"https://pith.science/api/pith-number/TNPWS2RGMTTJJB23R2XPM63EAA/graph.json","events_json":"https://pith.science/api/pith-number/TNPWS2RGMTTJJB23R2XPM63EAA/events.json","paper":"https://pith.science/paper/TNPWS2RG"},"agent_actions":{"view_html":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA","download_json":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA.json","view_paper":"https://pith.science/paper/TNPWS2RG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.16438&json=true","fetch_graph":"https://pith.science/api/pith-number/TNPWS2RGMTTJJB23R2XPM63EAA/graph.json","fetch_events":"https://pith.science/api/pith-number/TNPWS2RGMTTJJB23R2XPM63EAA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA/action/storage_attestation","attest_author":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA/action/author_attestation","sign_citation":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA/action/citation_signature","submit_replication":"https://pith.science/pith/TNPWS2RGMTTJJB23R2XPM63EAA/action/replication_record"}},"created_at":"2026-07-05T08:28:06.998346+00:00","updated_at":"2026-07-05T08:28:06.998346+00:00"}