{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RUOTT7YSHPJVJMC553ZCHQ4BVM","short_pith_number":"pith:RUOTT7YS","schema_version":"1.0","canonical_sha256":"8d1d39ff123bd354b05deef223c381ab3316bd094d431c5d2445c96434ffe558","source":{"kind":"arxiv","id":"2606.20225","version":1},"attestation_state":"computed","paper":{"title":"Actionable Activation Directions for Detecting and Mitigating Emergent Misalignment Across Language Model Families","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abdul Rafay Syed","submitted_at":"2026-06-18T13:39:59Z","abstract_excerpt":"Fine-tuning language models on insecure code induces emergent misalignment with poorly understood internal structure. We investigate whether this misalignment corresponds to a causally actionable activation-space direction shared across architectures. Across four instruction-tuned model families (Qwen2.5-1.5B, Gemma-2-2B, Llama-3.2-1B, Ministral-3-3B) finetuned identically, a difference-in-means direction achieves 99.6% separation of aligned and misaligned activations at each model's final layer. Causal steering by subtracting this direction reduces code spillover by 21-51 points, while a secu"},"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":"2606.20225","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-06-18T13:39:59Z","cross_cats_sorted":[],"title_canon_sha256":"51b15f9c640af4fbe46b548bad3a1f02d6895fdd13892598d80892fbc69914e7","abstract_canon_sha256":"bb316df2dd175a6dfa81fe6fcfb3cd9a4de663eba2a132fbed133285c1eb2385"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-19T16:13:06.335395Z","signature_b64":"6Ajv2FG8jycAYx0RcGYS5YRfYX1aBsU3j0yDpMzKwFrK4UNnHNpHTrBBPGenXwtbljRO9cmeo5XGlRWcjaLFCg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8d1d39ff123bd354b05deef223c381ab3316bd094d431c5d2445c96434ffe558","last_reissued_at":"2026-06-19T16:13:06.334970Z","signature_status":"signed_v1","first_computed_at":"2026-06-19T16:13:06.334970Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Actionable Activation Directions for Detecting and Mitigating Emergent Misalignment Across Language Model Families","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Abdul Rafay Syed","submitted_at":"2026-06-18T13:39:59Z","abstract_excerpt":"Fine-tuning language models on insecure code induces emergent misalignment with poorly understood internal structure. We investigate whether this misalignment corresponds to a causally actionable activation-space direction shared across architectures. Across four instruction-tuned model families (Qwen2.5-1.5B, Gemma-2-2B, Llama-3.2-1B, Ministral-3-3B) finetuned identically, a difference-in-means direction achieves 99.6% separation of aligned and misaligned activations at each model's final layer. Causal steering by subtracting this direction reduces code spillover by 21-51 points, while a secu"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.20225","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/2606.20225/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":"2606.20225","created_at":"2026-06-19T16:13:06.335034+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.20225v1","created_at":"2026-06-19T16:13:06.335034+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.20225","created_at":"2026-06-19T16:13:06.335034+00:00"},{"alias_kind":"pith_short_12","alias_value":"RUOTT7YSHPJV","created_at":"2026-06-19T16:13:06.335034+00:00"},{"alias_kind":"pith_short_16","alias_value":"RUOTT7YSHPJVJMC5","created_at":"2026-06-19T16:13:06.335034+00:00"},{"alias_kind":"pith_short_8","alias_value":"RUOTT7YS","created_at":"2026-06-19T16:13:06.335034+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/RUOTT7YSHPJVJMC553ZCHQ4BVM","json":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM.json","graph_json":"https://pith.science/api/pith-number/RUOTT7YSHPJVJMC553ZCHQ4BVM/graph.json","events_json":"https://pith.science/api/pith-number/RUOTT7YSHPJVJMC553ZCHQ4BVM/events.json","paper":"https://pith.science/paper/RUOTT7YS"},"agent_actions":{"view_html":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM","download_json":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM.json","view_paper":"https://pith.science/paper/RUOTT7YS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.20225&json=true","fetch_graph":"https://pith.science/api/pith-number/RUOTT7YSHPJVJMC553ZCHQ4BVM/graph.json","fetch_events":"https://pith.science/api/pith-number/RUOTT7YSHPJVJMC553ZCHQ4BVM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM/action/storage_attestation","attest_author":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM/action/author_attestation","sign_citation":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM/action/citation_signature","submit_replication":"https://pith.science/pith/RUOTT7YSHPJVJMC553ZCHQ4BVM/action/replication_record"}},"created_at":"2026-06-19T16:13:06.335034+00:00","updated_at":"2026-06-19T16:13:06.335034+00:00"}