{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:EDRWFDPE65L4W6IWUUB7YYGMNP","short_pith_number":"pith:EDRWFDPE","schema_version":"1.0","canonical_sha256":"20e3628de4f757cb7916a503fc60cc6bf94241c49a07c44401eedf5249ae4c87","source":{"kind":"arxiv","id":"2606.09068","version":1},"attestation_state":"computed","paper":{"title":"Emergent Misalignment Can Be Induced by Sycophancy and Reversed via Alignment Gating","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangtao Zhai, Han Wang, Kaiwei Zhang, Kaiyuan Ji, Qi Jia, Sicheng Wang, Xiangyang Zhu, Yuan Tian, Zongrui Wang","submitted_at":"2026-06-08T06:05:47Z","abstract_excerpt":"Prior work has shown that fine-tuning large language models on malicious or incorrect outputs in narrow domains can induce broad misalignment and harmful behavior, a phenomenon known as emergent misalignment. However, efficient methods for reversing such misalignment remain limited. In this work, we make two contributions. First, we identify sycophancy fine-tuning, i.e., training models to passively agree with users' incorrect opinions, as a previously underexplored driver of emergent misalignment, and show that it induces broad and severe misaligned behavior. Second, we propose Alignment Gati"},"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":"2606.09068","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-06-08T06:05:47Z","cross_cats_sorted":[],"title_canon_sha256":"3fae6bccd1d557924775fdb5144c960e58de6fb2ea64087567b8b289624cf852","abstract_canon_sha256":"aeaee072dd7440bc5399f88c7c6f8745ec19129c708579e0a2c52cd93e136c5a"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T02:07:56.740471Z","signature_b64":"0wN9GuwYYivMPwmc0ATfz1X55Sm3xpJvWPDjyOdYHqtI+7iPH+ZAPfblh72u9hQY0d5cgAbGNaG5OdMDI1QfBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"20e3628de4f757cb7916a503fc60cc6bf94241c49a07c44401eedf5249ae4c87","last_reissued_at":"2026-06-09T02:07:56.739579Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T02:07:56.739579Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Emergent Misalignment Can Be Induced by Sycophancy and Reversed via Alignment Gating","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Guangtao Zhai, Han Wang, Kaiwei Zhang, Kaiyuan Ji, Qi Jia, Sicheng Wang, Xiangyang Zhu, Yuan Tian, Zongrui Wang","submitted_at":"2026-06-08T06:05:47Z","abstract_excerpt":"Prior work has shown that fine-tuning large language models on malicious or incorrect outputs in narrow domains can induce broad misalignment and harmful behavior, a phenomenon known as emergent misalignment. However, efficient methods for reversing such misalignment remain limited. In this work, we make two contributions. First, we identify sycophancy fine-tuning, i.e., training models to passively agree with users' incorrect opinions, as a previously underexplored driver of emergent misalignment, and show that it induces broad and severe misaligned behavior. Second, we propose Alignment Gati"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.09068","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/2606.09068/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":"2606.09068","created_at":"2026-06-09T02:07:56.739727+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.09068v1","created_at":"2026-06-09T02:07:56.739727+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.09068","created_at":"2026-06-09T02:07:56.739727+00:00"},{"alias_kind":"pith_short_12","alias_value":"EDRWFDPE65L4","created_at":"2026-06-09T02:07:56.739727+00:00"},{"alias_kind":"pith_short_16","alias_value":"EDRWFDPE65L4W6IW","created_at":"2026-06-09T02:07:56.739727+00:00"},{"alias_kind":"pith_short_8","alias_value":"EDRWFDPE","created_at":"2026-06-09T02:07:56.739727+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/EDRWFDPE65L4W6IWUUB7YYGMNP","json":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP.json","graph_json":"https://pith.science/api/pith-number/EDRWFDPE65L4W6IWUUB7YYGMNP/graph.json","events_json":"https://pith.science/api/pith-number/EDRWFDPE65L4W6IWUUB7YYGMNP/events.json","paper":"https://pith.science/paper/EDRWFDPE"},"agent_actions":{"view_html":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP","download_json":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP.json","view_paper":"https://pith.science/paper/EDRWFDPE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.09068&json=true","fetch_graph":"https://pith.science/api/pith-number/EDRWFDPE65L4W6IWUUB7YYGMNP/graph.json","fetch_events":"https://pith.science/api/pith-number/EDRWFDPE65L4W6IWUUB7YYGMNP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP/action/storage_attestation","attest_author":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP/action/author_attestation","sign_citation":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP/action/citation_signature","submit_replication":"https://pith.science/pith/EDRWFDPE65L4W6IWUUB7YYGMNP/action/replication_record"}},"created_at":"2026-06-09T02:07:56.739727+00:00","updated_at":"2026-06-09T02:07:56.739727+00:00"}