{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:ZAIBYIG77FGA7N5PXTXLF45ZCM","short_pith_number":"pith:ZAIBYIG7","canonical_record":{"source":{"id":"2606.04160","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T19:23:46Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d02868e7637ca7479f8307e8854c9453c05884cd3fc822607556c557f6ccaa86","abstract_canon_sha256":"ffab7a63669d0f45b26f294ad129b4198e138b0566557cabbad7b04bb77ca819"},"schema_version":"1.0"},"canonical_sha256":"c8101c20dff94c0fb7afbceeb2f3b9133dff9435703ced08cd844234a0a98687","source":{"kind":"arxiv","id":"2606.04160","version":1},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.04160","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"arxiv_version","alias_value":"2606.04160v1","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04160","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_12","alias_value":"ZAIBYIG77FGA","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_16","alias_value":"ZAIBYIG77FGA7N5P","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_8","alias_value":"ZAIBYIG7","created_at":"2026-06-04T01:08:18Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:ZAIBYIG77FGA7N5PXTXLF45ZCM","target":"record","payload":{"canonical_record":{"source":{"id":"2606.04160","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T19:23:46Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"d02868e7637ca7479f8307e8854c9453c05884cd3fc822607556c557f6ccaa86","abstract_canon_sha256":"ffab7a63669d0f45b26f294ad129b4198e138b0566557cabbad7b04bb77ca819"},"schema_version":"1.0"},"canonical_sha256":"c8101c20dff94c0fb7afbceeb2f3b9133dff9435703ced08cd844234a0a98687","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-04T01:08:18.260874Z","signature_b64":"AAblsoJWzEKUEvgHARF2kOUb5D8Jnag1F/RSFsJ8a3lQByaNtz8+xtAvfH64BdUMdgCy9OcewTl2+w0vRuthBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c8101c20dff94c0fb7afbceeb2f3b9133dff9435703ced08cd844234a0a98687","last_reissued_at":"2026-06-04T01:08:18.259872Z","signature_status":"signed_v1","first_computed_at":"2026-06-04T01:08:18.259872Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2606.04160","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":"2026-06-04T01:08:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"J6GxwTgdFHtBvtKiUEM80pZh8nCvl/Y6u5N2ZA5TPHdeD/hW34ZLpHnBhTyzpOl5MsAcuhAY10t8+DphKkM/Cg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-02T12:53:39.240997Z"},"content_sha256":"34e2d55a7a8055b727a76c983eb1312e87d789f75b5f053c7ae5308ba2f7b309","schema_version":"1.0","event_id":"sha256:34e2d55a7a8055b727a76c983eb1312e87d789f75b5f053c7ae5308ba2f7b309"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:ZAIBYIG77FGA7N5PXTXLF45ZCM","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Expert-Aware Refusal Steering","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Anna C. Marbut, Daniel R. Olson, Travis J. Wheeler","submitted_at":"2026-06-02T19:23:46Z","abstract_excerpt":"Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse to respond to harmful or disallowed requests. Recent work has shown that a steering vector can be applied to a dense LLM during inference to effectively suppress refusal behavior, inducing response to harmful requests. We extend this refusal steering method to three open-source Mixture-of-Experts (MoE) LLMs and find that steering performance is uninhibited by the complex routing patterns inherent to the MoE architecture. We then propose two expert-aware refusal steering methods th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04160","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/2606.04160/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":"2026-06-04T01:08:18Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"7DZFFPC6H/Atw5QT/g4wIl+FgZ65qdydzDs5ZW7Hke0y+LdXarZMqR4uvrWdCNaX9D9q5iinQMkBaYaNVSMOAQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-02T12:53:39.241408Z"},"content_sha256":"2cd1b07658a1076db57aa6f8ab1831f05d710154b389f7b2173d7c69b1b670d7","schema_version":"1.0","event_id":"sha256:2cd1b07658a1076db57aa6f8ab1831f05d710154b389f7b2173d7c69b1b670d7"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/bundle.json","state_url":"https://pith.science/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/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":"2026-07-02T12:53:39Z","links":{"resolver":"https://pith.science/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM","bundle":"https://pith.science/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/bundle.json","state":"https://pith.science/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/state.json","well_known_bundle":"https://pith.science/.well-known/pith/ZAIBYIG77FGA7N5PXTXLF45ZCM/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:ZAIBYIG77FGA7N5PXTXLF45ZCM","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":"ffab7a63669d0f45b26f294ad129b4198e138b0566557cabbad7b04bb77ca819","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T19:23:46Z","title_canon_sha256":"d02868e7637ca7479f8307e8854c9453c05884cd3fc822607556c557f6ccaa86"},"schema_version":"1.0","source":{"id":"2606.04160","kind":"arxiv","version":1}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2606.04160","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"arxiv_version","alias_value":"2606.04160v1","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.04160","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_12","alias_value":"ZAIBYIG77FGA","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_16","alias_value":"ZAIBYIG77FGA7N5P","created_at":"2026-06-04T01:08:18Z"},{"alias_kind":"pith_short_8","alias_value":"ZAIBYIG7","created_at":"2026-06-04T01:08:18Z"}],"graph_snapshots":[{"event_id":"sha256:2cd1b07658a1076db57aa6f8ab1831f05d710154b389f7b2173d7c69b1b670d7","target":"graph","created_at":"2026-06-04T01:08:18Z","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/2606.04160/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Safety alignment in instruction-tuned large language models (LLMs) depends on a model's ability to reliably refuse to respond to harmful or disallowed requests. Recent work has shown that a steering vector can be applied to a dense LLM during inference to effectively suppress refusal behavior, inducing response to harmful requests. We extend this refusal steering method to three open-source Mixture-of-Experts (MoE) LLMs and find that steering performance is uninhibited by the complex routing patterns inherent to the MoE architecture. We then propose two expert-aware refusal steering methods th","authors_text":"Anna C. Marbut, Daniel R. Olson, Travis J. Wheeler","cross_cats":["cs.LG"],"headline":"","license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T19:23:46Z","title":"Expert-Aware Refusal Steering"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.04160","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":"sha256:34e2d55a7a8055b727a76c983eb1312e87d789f75b5f053c7ae5308ba2f7b309","target":"record","created_at":"2026-06-04T01:08:18Z","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":"ffab7a63669d0f45b26f294ad129b4198e138b0566557cabbad7b04bb77ca819","cross_cats_sorted":["cs.LG"],"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2026-06-02T19:23:46Z","title_canon_sha256":"d02868e7637ca7479f8307e8854c9453c05884cd3fc822607556c557f6ccaa86"},"schema_version":"1.0","source":{"id":"2606.04160","kind":"arxiv","version":1}},"canonical_sha256":"c8101c20dff94c0fb7afbceeb2f3b9133dff9435703ced08cd844234a0a98687","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"c8101c20dff94c0fb7afbceeb2f3b9133dff9435703ced08cd844234a0a98687","first_computed_at":"2026-06-04T01:08:18.259872Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-04T01:08:18.259872Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"AAblsoJWzEKUEvgHARF2kOUb5D8Jnag1F/RSFsJ8a3lQByaNtz8+xtAvfH64BdUMdgCy9OcewTl2+w0vRuthBQ==","signature_status":"signed_v1","signed_at":"2026-06-04T01:08:18.260874Z","signed_message":"canonical_sha256_bytes"},"source_id":"2606.04160","source_kind":"arxiv","source_version":1}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:34e2d55a7a8055b727a76c983eb1312e87d789f75b5f053c7ae5308ba2f7b309","sha256:2cd1b07658a1076db57aa6f8ab1831f05d710154b389f7b2173d7c69b1b670d7"],"state_sha256":"6b6ee3379263a7baa1e3ffe5fc9e1ac0fa49e8b120cfe91160bd6be151eeb73a"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"gNFr+O3NP3QOKJmYAkWlspfSRx05cUptOFkZLHLwXldAnUsFpX96hBW9IRlJm7JqfaFojuPedxwg2lHpMlumCA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-02T12:53:39.243392Z","bundle_sha256":"0779d9a0e6d3689a9bdfce5aacb3d305edf7ac40cc00a0cefe59e36922ff3e15"}}