{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:M2YG2BONZ4YXVWSJS32IDTXT5O","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":"a12f503049a943cedcb626d29d9779ad1c7d54392423308750220cd10bb22559","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-13T09:39:54Z","title_canon_sha256":"cc860aeae99a20b07163f62ce61c6d9b7ad54277edbea8d3b44f18214531feb8"},"schema_version":"1.0","source":{"id":"2506.11618","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2506.11618","created_at":"2026-07-05T11:24:53Z"},{"alias_kind":"arxiv_version","alias_value":"2506.11618v2","created_at":"2026-07-05T11:24:53Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.11618","created_at":"2026-07-05T11:24:53Z"},{"alias_kind":"pith_short_12","alias_value":"M2YG2BONZ4YX","created_at":"2026-07-05T11:24:53Z"},{"alias_kind":"pith_short_16","alias_value":"M2YG2BONZ4YXVWSJ","created_at":"2026-07-05T11:24:53Z"},{"alias_kind":"pith_short_8","alias_value":"M2YG2BON","created_at":"2026-07-05T11:24:53Z"}],"graph_snapshots":[{"event_id":"sha256:f0e49f5067ae6c0fcf4849f0a72a0a9293db5f69fbab24304439adae4facb4fc","target":"graph","created_at":"2026-07-05T11:24:53Z","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/2506.11618/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Fine-tuning large language models on narrow datasets can cause them to develop broadly misaligned behaviours: a phenomena known as emergent misalignment. However, the mechanisms underlying this misalignment, and why it generalizes beyond the training domain, are poorly understood, demonstrating critical gaps in our knowledge of model alignment. In this work, we train and study a minimal model organism which uses just 9 rank-1 adapters to emergently misalign Qwen2.5-14B-Instruct. Studying this, we find that different emergently misaligned models converge to similar representations of misalignme","authors_text":"Anna Soligo, Edward Turner, Neel Nanda, Senthooran Rajamanoharan","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-13T09:39:54Z","title":"Convergent Linear Representations of Emergent Misalignment"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.11618","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":"sha256:5a8c821d09ae1159923b1c29cc4f114bc51edf913ff766b2cf295e43ef3f7bb5","target":"record","created_at":"2026-07-05T11:24:53Z","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":"a12f503049a943cedcb626d29d9779ad1c7d54392423308750220cd10bb22559","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-06-13T09:39:54Z","title_canon_sha256":"cc860aeae99a20b07163f62ce61c6d9b7ad54277edbea8d3b44f18214531feb8"},"schema_version":"1.0","source":{"id":"2506.11618","kind":"arxiv","version":2}},"canonical_sha256":"66b06d05cdcf317ada4996f481cef3ebaaaff6bc785ab32db9279a6af82d6439","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"66b06d05cdcf317ada4996f481cef3ebaaaff6bc785ab32db9279a6af82d6439","first_computed_at":"2026-07-05T11:24:53.517116Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T11:24:53.517116Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"c/MEyZBulP+33KvzfdllrwVV1xQBHCgm4QRwwOk1aMryDnr6vCrMTeZ6c580o4AqreF5qz4+WhNYmpJuK9cqAg==","signature_status":"signed_v1","signed_at":"2026-07-05T11:24:53.517625Z","signed_message":"canonical_sha256_bytes"},"source_id":"2506.11618","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:5a8c821d09ae1159923b1c29cc4f114bc51edf913ff766b2cf295e43ef3f7bb5","sha256:f0e49f5067ae6c0fcf4849f0a72a0a9293db5f69fbab24304439adae4facb4fc"],"state_sha256":"761abf549a073eb46beaeeb5bfedaddfdb75384c9a5e04616d8e026da68317de"}