{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:FDD53OXA2CLX6NZ74PYVVQIFQO","short_pith_number":"pith:FDD53OXA","schema_version":"1.0","canonical_sha256":"28c7ddbae0d0977f373fe3f15ac1058387df07eab447f83c40cce2a59ab2e062","source":{"kind":"arxiv","id":"2402.10588","version":4},"attestation_state":"computed","paper":{"title":"Do Llamas Work in English? On the Latent Language of Multilingual Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Chris Wendler, Giovanni Monea, Robert West, Veniamin Veselovsky","submitted_at":"2024-02-16T11:21:28Z","abstract_excerpt":"We ask whether multilingual language models trained on unbalanced, English-dominated corpora use English as an internal pivot language -- a question of key importance for understanding how language models function and the origins of linguistic bias. Focusing on the Llama-2 family of transformer models, our study uses carefully constructed non-English prompts with a unique correct single-token continuation. From layer to layer, transformers gradually map an input embedding of the final prompt token to an output embedding from which next-token probabilities are computed. Tracking intermediate em"},"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":"2402.10588","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-16T11:21:28Z","cross_cats_sorted":["cs.CY"],"title_canon_sha256":"4cdbcf5b86cf962e0ba0bc2ed84abd90078aaf9217c877aa94931ac11e13292b","abstract_canon_sha256":"0ca23c19f9f3591686c4970bc2e644035699cb66027d435216d99b42ff1333a2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:30:08.795526Z","signature_b64":"RjMIgG1swbaQj5h4f8gpLX9Af0lbgdLix7iNJ14Ci9KZmJ3dTJshMwhXwoAVHo1ueGzMQaujNERZmQmYGvAMBA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"28c7ddbae0d0977f373fe3f15ac1058387df07eab447f83c40cce2a59ab2e062","last_reissued_at":"2026-07-05T08:30:08.795041Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:30:08.795041Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Do Llamas Work in English? On the Latent Language of Multilingual Transformers","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CY"],"primary_cat":"cs.CL","authors_text":"Chris Wendler, Giovanni Monea, Robert West, Veniamin Veselovsky","submitted_at":"2024-02-16T11:21:28Z","abstract_excerpt":"We ask whether multilingual language models trained on unbalanced, English-dominated corpora use English as an internal pivot language -- a question of key importance for understanding how language models function and the origins of linguistic bias. Focusing on the Llama-2 family of transformer models, our study uses carefully constructed non-English prompts with a unique correct single-token continuation. From layer to layer, transformers gradually map an input embedding of the final prompt token to an output embedding from which next-token probabilities are computed. Tracking intermediate em"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.10588","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/2402.10588/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":"2402.10588","created_at":"2026-07-05T08:30:08.795093+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.10588v4","created_at":"2026-07-05T08:30:08.795093+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.10588","created_at":"2026-07-05T08:30:08.795093+00:00"},{"alias_kind":"pith_short_12","alias_value":"FDD53OXA2CLX","created_at":"2026-07-05T08:30:08.795093+00:00"},{"alias_kind":"pith_short_16","alias_value":"FDD53OXA2CLX6NZ7","created_at":"2026-07-05T08:30:08.795093+00:00"},{"alias_kind":"pith_short_8","alias_value":"FDD53OXA","created_at":"2026-07-05T08:30:08.795093+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2604.27401","citing_title":"Perturbation Probing: A Two-Pass-per-Prompt Diagnostic for FFN Behavioral Circuits in Aligned LLMs","ref_index":26,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO","json":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO.json","graph_json":"https://pith.science/api/pith-number/FDD53OXA2CLX6NZ74PYVVQIFQO/graph.json","events_json":"https://pith.science/api/pith-number/FDD53OXA2CLX6NZ74PYVVQIFQO/events.json","paper":"https://pith.science/paper/FDD53OXA"},"agent_actions":{"view_html":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO","download_json":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO.json","view_paper":"https://pith.science/paper/FDD53OXA","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.10588&json=true","fetch_graph":"https://pith.science/api/pith-number/FDD53OXA2CLX6NZ74PYVVQIFQO/graph.json","fetch_events":"https://pith.science/api/pith-number/FDD53OXA2CLX6NZ74PYVVQIFQO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO/action/storage_attestation","attest_author":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO/action/author_attestation","sign_citation":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO/action/citation_signature","submit_replication":"https://pith.science/pith/FDD53OXA2CLX6NZ74PYVVQIFQO/action/replication_record"}},"created_at":"2026-07-05T08:30:08.795093+00:00","updated_at":"2026-07-05T08:30:08.795093+00:00"}