{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:A7PA4CSBZ2DCTXMHZHY7CGN7P4","short_pith_number":"pith:A7PA4CSB","schema_version":"1.0","canonical_sha256":"07de0e0a41ce8629dd87c9f1f119bf7f086fafdeb55713fecb399e67684719bb","source":{"kind":"arxiv","id":"2506.15415","version":1},"attestation_state":"computed","paper":{"title":"Targeted Lexical Injection: Unlocking Latent Cross-Lingual Alignment in Lugha-Llama via Early-Layer LoRA Fine-Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Stanley Ngugi","submitted_at":"2025-06-18T12:35:53Z","abstract_excerpt":"Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their performance in low-resource languages (LRLs), such as Swahili, often lags due to data scarcity and underrepresentation in pre-training. A key challenge is achieving robust cross-lingual lexical alignment, crucial for tasks like translation and cross-lingual information retrieval. This paper introduces Targeted Lexical Injection (TLI), a novel and efficient fine-tuning approach. We first demonstrate that Lugha-Llama-8B-wura, a Swahili-centric LLM, exhibits strong, near-perfect lexical alignment for Swahili-English"},"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":"2506.15415","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-06-18T12:35:53Z","cross_cats_sorted":[],"title_canon_sha256":"ffcbdc1d7505039c216711a8cf3a08f86cd3bb8cf6d706324146438ab2d8ea6f","abstract_canon_sha256":"0faeffebaa48f3481288094559652f149d39f88d21b2ca10f35145ed6df4428d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:23:42.854195Z","signature_b64":"vssn8I2f+/l4nidqYZsGqgOmXmqODT5+RZMbf6MfYuhKgUEIfrWE/wf1K2DWXjd3XZljKC7EABwjK8RxNtimCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"07de0e0a41ce8629dd87c9f1f119bf7f086fafdeb55713fecb399e67684719bb","last_reissued_at":"2026-07-05T11:23:42.853487Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:23:42.853487Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Targeted Lexical Injection: Unlocking Latent Cross-Lingual Alignment in Lugha-Llama via Early-Layer LoRA Fine-Tuning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Stanley Ngugi","submitted_at":"2025-06-18T12:35:53Z","abstract_excerpt":"Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their performance in low-resource languages (LRLs), such as Swahili, often lags due to data scarcity and underrepresentation in pre-training. A key challenge is achieving robust cross-lingual lexical alignment, crucial for tasks like translation and cross-lingual information retrieval. This paper introduces Targeted Lexical Injection (TLI), a novel and efficient fine-tuning approach. We first demonstrate that Lugha-Llama-8B-wura, a Swahili-centric LLM, exhibits strong, near-perfect lexical alignment for Swahili-English"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.15415","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/2506.15415/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":"2506.15415","created_at":"2026-07-05T11:23:42.853578+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.15415v1","created_at":"2026-07-05T11:23:42.853578+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.15415","created_at":"2026-07-05T11:23:42.853578+00:00"},{"alias_kind":"pith_short_12","alias_value":"A7PA4CSBZ2DC","created_at":"2026-07-05T11:23:42.853578+00:00"},{"alias_kind":"pith_short_16","alias_value":"A7PA4CSBZ2DCTXMH","created_at":"2026-07-05T11:23:42.853578+00:00"},{"alias_kind":"pith_short_8","alias_value":"A7PA4CSB","created_at":"2026-07-05T11:23:42.853578+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/A7PA4CSBZ2DCTXMHZHY7CGN7P4","json":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4.json","graph_json":"https://pith.science/api/pith-number/A7PA4CSBZ2DCTXMHZHY7CGN7P4/graph.json","events_json":"https://pith.science/api/pith-number/A7PA4CSBZ2DCTXMHZHY7CGN7P4/events.json","paper":"https://pith.science/paper/A7PA4CSB"},"agent_actions":{"view_html":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4","download_json":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4.json","view_paper":"https://pith.science/paper/A7PA4CSB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.15415&json=true","fetch_graph":"https://pith.science/api/pith-number/A7PA4CSBZ2DCTXMHZHY7CGN7P4/graph.json","fetch_events":"https://pith.science/api/pith-number/A7PA4CSBZ2DCTXMHZHY7CGN7P4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4/action/storage_attestation","attest_author":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4/action/author_attestation","sign_citation":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4/action/citation_signature","submit_replication":"https://pith.science/pith/A7PA4CSBZ2DCTXMHZHY7CGN7P4/action/replication_record"}},"created_at":"2026-07-05T11:23:42.853578+00:00","updated_at":"2026-07-05T11:23:42.853578+00:00"}