{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PGKQUGQLU7ROQR76FZW25NP22R","short_pith_number":"pith:PGKQUGQL","schema_version":"1.0","canonical_sha256":"79950a1a0ba7e2e847fe2e6daeb5fad4668326c7d12464453c855b36468dc6a3","source":{"kind":"arxiv","id":"2505.03501","version":1},"attestation_state":"computed","paper":{"title":"BadLingual: A Novel Lingual-Backdoor Attack against Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Guowen Xu, Hongwei Li, Kangjie Chen, Qingchuan Zhao, Rui Zhang, Tianwei Zhang, Wenbo Jiang, Zihan Wang","submitted_at":"2025-05-06T13:07:57Z","abstract_excerpt":"In this paper, we present a new form of backdoor attack against Large Language Models (LLMs): lingual-backdoor attacks. The key novelty of lingual-backdoor attacks is that the language itself serves as the trigger to hijack the infected LLMs to generate inflammatory speech. They enable the precise targeting of a specific language-speaking group, exacerbating racial discrimination by malicious entities. We first implement a baseline lingual-backdoor attack, which is carried out by poisoning a set of training data for specific downstream tasks through translation into the trigger language. Howev"},"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":"2505.03501","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CR","submitted_at":"2025-05-06T13:07:57Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6f2577e08158f2a85bff3621e432b8bf8caaf59af77e77f2b36c77cc1570388f","abstract_canon_sha256":"55842d0fbf533be44c1f84d02242da537ed978807e1d91232a34537997e5b921"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:59:17.160915Z","signature_b64":"iW9quJZNXD6zRaAe0G1o4tUUZ9ht5puxJBOyfgHNt70pa9v0pOrnb/x5WsYqd36WhUPnxfZL73DecuDxrzBZCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"79950a1a0ba7e2e847fe2e6daeb5fad4668326c7d12464453c855b36468dc6a3","last_reissued_at":"2026-07-05T10:59:17.160434Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:59:17.160434Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"BadLingual: A Novel Lingual-Backdoor Attack against Large Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CR","authors_text":"Guowen Xu, Hongwei Li, Kangjie Chen, Qingchuan Zhao, Rui Zhang, Tianwei Zhang, Wenbo Jiang, Zihan Wang","submitted_at":"2025-05-06T13:07:57Z","abstract_excerpt":"In this paper, we present a new form of backdoor attack against Large Language Models (LLMs): lingual-backdoor attacks. The key novelty of lingual-backdoor attacks is that the language itself serves as the trigger to hijack the infected LLMs to generate inflammatory speech. They enable the precise targeting of a specific language-speaking group, exacerbating racial discrimination by malicious entities. We first implement a baseline lingual-backdoor attack, which is carried out by poisoning a set of training data for specific downstream tasks through translation into the trigger language. Howev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.03501","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/2505.03501/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":"2505.03501","created_at":"2026-07-05T10:59:17.160492+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.03501v1","created_at":"2026-07-05T10:59:17.160492+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.03501","created_at":"2026-07-05T10:59:17.160492+00:00"},{"alias_kind":"pith_short_12","alias_value":"PGKQUGQLU7RO","created_at":"2026-07-05T10:59:17.160492+00:00"},{"alias_kind":"pith_short_16","alias_value":"PGKQUGQLU7ROQR76","created_at":"2026-07-05T10:59:17.160492+00:00"},{"alias_kind":"pith_short_8","alias_value":"PGKQUGQL","created_at":"2026-07-05T10:59:17.160492+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.25073","citing_title":"Security in the Fine-Tuning Lifecycle of Large Language Models: Threats, Defenses,Evaluation, and Future Directions","ref_index":75,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R","json":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R.json","graph_json":"https://pith.science/api/pith-number/PGKQUGQLU7ROQR76FZW25NP22R/graph.json","events_json":"https://pith.science/api/pith-number/PGKQUGQLU7ROQR76FZW25NP22R/events.json","paper":"https://pith.science/paper/PGKQUGQL"},"agent_actions":{"view_html":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R","download_json":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R.json","view_paper":"https://pith.science/paper/PGKQUGQL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.03501&json=true","fetch_graph":"https://pith.science/api/pith-number/PGKQUGQLU7ROQR76FZW25NP22R/graph.json","fetch_events":"https://pith.science/api/pith-number/PGKQUGQLU7ROQR76FZW25NP22R/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R/action/storage_attestation","attest_author":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R/action/author_attestation","sign_citation":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R/action/citation_signature","submit_replication":"https://pith.science/pith/PGKQUGQLU7ROQR76FZW25NP22R/action/replication_record"}},"created_at":"2026-07-05T10:59:17.160492+00:00","updated_at":"2026-07-05T10:59:17.160492+00:00"}