{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:HOT3FGBNOXERT3FMHUCAVBECEY","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"d92a0b2cd85ec3cb0f0c7380f93dfe4b3414395fad47116743cec116bf542f24","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-16T19:44:01Z","title_canon_sha256":"b78b8308835d0921483ea3c0ceaee7d2e6b1aa70667361e8b3840d5ddae3fcba"},"schema_version":"1.0","source":{"id":"2502.11244","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2502.11244","created_at":"2026-07-05T11:57:34Z"},{"alias_kind":"arxiv_version","alias_value":"2502.11244v2","created_at":"2026-07-05T11:57:34Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2502.11244","created_at":"2026-07-05T11:57:34Z"},{"alias_kind":"pith_short_12","alias_value":"HOT3FGBNOXER","created_at":"2026-07-05T11:57:34Z"},{"alias_kind":"pith_short_16","alias_value":"HOT3FGBNOXERT3FM","created_at":"2026-07-05T11:57:34Z"},{"alias_kind":"pith_short_8","alias_value":"HOT3FGBN","created_at":"2026-07-05T11:57:34Z"}],"graph_snapshots":[{"event_id":"sha256:03dc8ccd271da9bcbdea81c1a8dac6def085e361933af952c59e1c1a8bfffc23","target":"graph","created_at":"2026-07-05T11:57:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2502.11244/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Ensuring consistent safety across multiple languages remains a significant challenge for large language models (LLMs). We introduce Soteria, a lightweight yet powerful strategy that locates and minimally adjusts the \"functional heads\" most responsible for harmful content generation in each language. By altering only a fraction of parameters, Soteria drastically reduces policy violations without sacrificing overall model performance, even in low-resource settings. To rigorously evaluate our approach, we also present XThreatBench, a specialized multilingual dataset capturing fine-grained harmful","authors_text":"Animesh Mukherjee, Pratyush Chatterjee, Rima Hazra, Sayan Layek, Somnath Banerjee","cross_cats":["cs.AI"],"headline":"","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-16T19:44:01Z","title":"Soteria: Language-Specific Functional Parameter Steering for Multilingual Safety Alignment"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2502.11244","kind":"arxiv","version":2},"verdict":{"created_at":null,"id":null,"model_set":{},"one_line_summary":"","pipeline_version":null,"pith_extraction_headline":"","strongest_claim":"","weakest_assumption":""}},"verdict_id":null}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:28ec5b0a208924aa080a71ddf747ddba94ac67e1e054afbf3e249304f419df91","target":"record","created_at":"2026-07-05T11:57:34Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"d92a0b2cd85ec3cb0f0c7380f93dfe4b3414395fad47116743cec116bf542f24","cross_cats_sorted":["cs.AI"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2025-02-16T19:44:01Z","title_canon_sha256":"b78b8308835d0921483ea3c0ceaee7d2e6b1aa70667361e8b3840d5ddae3fcba"},"schema_version":"1.0","source":{"id":"2502.11244","kind":"arxiv","version":2}},"canonical_sha256":"3ba7b2982d75c919ecac3d040a8482260e82705446f0ee94efb2c16eb63d597b","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"3ba7b2982d75c919ecac3d040a8482260e82705446f0ee94efb2c16eb63d597b","first_computed_at":"2026-07-05T11:57:34.619452Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:57:34.619452Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"x2Jt9H+Kut+GpjLcqF4K116P2CEWHLqwN9mFyHO0FkStuF4YWAfUI/bECJ42ce3epu2TlbtBQNZ+J2KKQ8iKDA==","signature_status":"signed_v1","signed_at":"2026-07-05T11:57:34.619920Z","signed_message":"canonical_sha256_bytes"},"source_id":"2502.11244","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:28ec5b0a208924aa080a71ddf747ddba94ac67e1e054afbf3e249304f419df91","sha256:03dc8ccd271da9bcbdea81c1a8dac6def085e361933af952c59e1c1a8bfffc23"],"state_sha256":"6eac0e8d30d2f196e373795628e68606b5adb31afb2876e65a639863f927b068"}