{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2025:PD2T6QDGF5YJ4CADJUOIOCSJGH","short_pith_number":"pith:PD2T6QDG","canonical_record":{"source":{"id":"2501.08145","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T14:23:18Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e076fae057e939a2f7e8ea983e272a9bd569672f1da7d31ece8e614001c1d0c1","abstract_canon_sha256":"eae5c645db01bcc81e478afa0abb46ab1ea8295da752a535f9f9a30aa548de34"},"schema_version":"1.0"},"canonical_sha256":"78f53f40662f709e08034d1c870a4931c7a8dd258810cc937fab80348c374b92","source":{"kind":"arxiv","id":"2501.08145","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.08145","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"arxiv_version","alias_value":"2501.08145v1","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.08145","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_12","alias_value":"PD2T6QDGF5YJ","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_16","alias_value":"PD2T6QDGF5YJ4CAD","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_8","alias_value":"PD2T6QDG","created_at":"2026-07-05T10:00:56Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2025:PD2T6QDGF5YJ4CADJUOIOCSJGH","target":"record","payload":{"canonical_record":{"source":{"id":"2501.08145","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T14:23:18Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e076fae057e939a2f7e8ea983e272a9bd569672f1da7d31ece8e614001c1d0c1","abstract_canon_sha256":"eae5c645db01bcc81e478afa0abb46ab1ea8295da752a535f9f9a30aa548de34"},"schema_version":"1.0"},"canonical_sha256":"78f53f40662f709e08034d1c870a4931c7a8dd258810cc937fab80348c374b92","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:00:56.599579Z","signature_b64":"wIbI9zWSP3lkGpL63feaQxITh9fRKgSItJrI2snZL6Mruivl8gN9AGhk0H0XXIbVh7HNHVqSyEc6u0vNx28IAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"78f53f40662f709e08034d1c870a4931c7a8dd258810cc937fab80348c374b92","last_reissued_at":"2026-07-05T10:00:56.599118Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:00:56.599118Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2501.08145","source_version":1,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:00:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"8y+ZC6TVD6aNo6sWLfcBlG+2magpYftRVjU5p2fnWwC1cmzHNPl7KjsGMCpWk44IMZISJBgmVQe15cIbFoBkCA==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T15:54:48.097486Z"},"content_sha256":"869d937fca47700f268e5229ff78f5a2842b5adabf8352395389dc6bed81d765","schema_version":"1.0","event_id":"sha256:869d937fca47700f268e5229ff78f5a2842b5adabf8352395389dc6bed81d765"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2025:PD2T6QDGF5YJ4CADJUOIOCSJGH","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Refusal Behavior in Large Language Models: A Nonlinear Perspective","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Achim Schilling, Andreas Maier, Fabian Hildebrandt, Patrick Krauss","submitted_at":"2025-01-14T14:23:18Z","abstract_excerpt":"Refusal behavior in large language models (LLMs) enables them to decline responding to harmful, unethical, or inappropriate prompts, ensuring alignment with ethical standards. This paper investigates refusal behavior across six LLMs from three architectural families. We challenge the assumption of refusal as a linear phenomenon by employing dimensionality reduction techniques, including PCA, t-SNE, and UMAP. Our results reveal that refusal mechanisms exhibit nonlinear, multidimensional characteristics that vary by model architecture and layer. These findings highlight the need for nonlinear in"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.08145","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/2501.08145/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"},"verdict_id":null},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-07-05T10:00:56Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"qcQgED9eGeuhINyqGclpemGP7c2CGoF/JQqlfYu1I6in9j60EFTRxTQcEtvN4W2rt6dPUDR7+iUEiIVReqpPCQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-07T15:54:48.097867Z"},"content_sha256":"784d6f1fe21d6e4b53e28f8164e927bd5e40b8e5e5e00419e085025512184b69","schema_version":"1.0","event_id":"sha256:784d6f1fe21d6e4b53e28f8164e927bd5e40b8e5e5e00419e085025512184b69"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/bundle.json","state_url":"https://pith.science/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-07-07T15:54:48Z","links":{"resolver":"https://pith.science/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH","bundle":"https://pith.science/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/bundle.json","state":"https://pith.science/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/state.json","well_known_bundle":"https://pith.science/.well-known/pith/PD2T6QDGF5YJ4CADJUOIOCSJGH/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2025:PD2T6QDGF5YJ4CADJUOIOCSJGH","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":"eae5c645db01bcc81e478afa0abb46ab1ea8295da752a535f9f9a30aa548de34","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T14:23:18Z","title_canon_sha256":"e076fae057e939a2f7e8ea983e272a9bd569672f1da7d31ece8e614001c1d0c1"},"schema_version":"1.0","source":{"id":"2501.08145","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2501.08145","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"arxiv_version","alias_value":"2501.08145v1","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.08145","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_12","alias_value":"PD2T6QDGF5YJ","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_16","alias_value":"PD2T6QDGF5YJ4CAD","created_at":"2026-07-05T10:00:56Z"},{"alias_kind":"pith_short_8","alias_value":"PD2T6QDG","created_at":"2026-07-05T10:00:56Z"}],"graph_snapshots":[{"event_id":"sha256:784d6f1fe21d6e4b53e28f8164e927bd5e40b8e5e5e00419e085025512184b69","target":"graph","created_at":"2026-07-05T10:00:56Z","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/2501.08145/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Refusal behavior in large language models (LLMs) enables them to decline responding to harmful, unethical, or inappropriate prompts, ensuring alignment with ethical standards. This paper investigates refusal behavior across six LLMs from three architectural families. We challenge the assumption of refusal as a linear phenomenon by employing dimensionality reduction techniques, including PCA, t-SNE, and UMAP. Our results reveal that refusal mechanisms exhibit nonlinear, multidimensional characteristics that vary by model architecture and layer. These findings highlight the need for nonlinear in","authors_text":"Achim Schilling, Andreas Maier, Fabian Hildebrandt, Patrick Krauss","cross_cats":["cs.AI"],"headline":"","license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T14:23:18Z","title":"Refusal Behavior in Large Language Models: A Nonlinear Perspective"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.08145","kind":"arxiv","version":1},"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":"sha256:869d937fca47700f268e5229ff78f5a2842b5adabf8352395389dc6bed81d765","target":"record","created_at":"2026-07-05T10:00:56Z","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":"eae5c645db01bcc81e478afa0abb46ab1ea8295da752a535f9f9a30aa548de34","cross_cats_sorted":["cs.AI"],"license":"http://creativecommons.org/licenses/by-nc-nd/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T14:23:18Z","title_canon_sha256":"e076fae057e939a2f7e8ea983e272a9bd569672f1da7d31ece8e614001c1d0c1"},"schema_version":"1.0","source":{"id":"2501.08145","kind":"arxiv","version":1}},"canonical_sha256":"78f53f40662f709e08034d1c870a4931c7a8dd258810cc937fab80348c374b92","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"78f53f40662f709e08034d1c870a4931c7a8dd258810cc937fab80348c374b92","first_computed_at":"2026-07-05T10:00:56.599118Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-07-05T10:00:56.599118Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"wIbI9zWSP3lkGpL63feaQxITh9fRKgSItJrI2snZL6Mruivl8gN9AGhk0H0XXIbVh7HNHVqSyEc6u0vNx28IAg==","signature_status":"signed_v1","signed_at":"2026-07-05T10:00:56.599579Z","signed_message":"canonical_sha256_bytes"},"source_id":"2501.08145","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:869d937fca47700f268e5229ff78f5a2842b5adabf8352395389dc6bed81d765","sha256:784d6f1fe21d6e4b53e28f8164e927bd5e40b8e5e5e00419e085025512184b69"],"state_sha256":"ed0b3be2d7a299f97d481845273abbe0fad9b9f77595453278cb9a4af36ce8a0"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"AMQhI7pEHer0B950MI0FteCe1JuB4nG+EKKFgjY21zJBbyMc/R+MzNBf19DpwsHLomTaFzsOkcCwKAWlL2a1CA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-07T15:54:48.099843Z","bundle_sha256":"7df5e5c824dfa67a0170c9919403c8e07d0db6bff49c9b50edf4485f4f9e4654"}}