{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:W3AYUNLZEN3MUG2TPF2KH65SME","short_pith_number":"pith:W3AYUNLZ","canonical_record":{"source":{"id":"2604.04385","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-06T03:20:37Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2535a2bd1b60beef235b319f09b2c08e0dc47b227dca504ee3bdfb81f47b9caf","abstract_canon_sha256":"f522f2b99c1fb0ffb7e53ab70c8fefe8c3473976e4c4ffa1703bcdf5cbd639ec"},"schema_version":"1.0"},"canonical_sha256":"b6c18a35792376ca1b537974a3fbb2611876e15fd7f7ed7276e3dde75a33cdf6","source":{"kind":"arxiv","id":"2604.04385","version":5},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.04385","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"arxiv_version","alias_value":"2604.04385v5","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.04385","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_12","alias_value":"W3AYUNLZEN3M","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_16","alias_value":"W3AYUNLZEN3MUG2T","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_8","alias_value":"W3AYUNLZ","created_at":"2026-06-30T02:17:20Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:W3AYUNLZEN3MUG2TPF2KH65SME","target":"record","payload":{"canonical_record":{"source":{"id":"2604.04385","kind":"arxiv","version":5},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-06T03:20:37Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"2535a2bd1b60beef235b319f09b2c08e0dc47b227dca504ee3bdfb81f47b9caf","abstract_canon_sha256":"f522f2b99c1fb0ffb7e53ab70c8fefe8c3473976e4c4ffa1703bcdf5cbd639ec"},"schema_version":"1.0"},"canonical_sha256":"b6c18a35792376ca1b537974a3fbb2611876e15fd7f7ed7276e3dde75a33cdf6","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-30T02:17:20.139021Z","signature_b64":"nYRdF2mIY5V/2hQo8c+gZyRvrMxbR+V4aLkPY6LkP31HQsxmyFBOpzSfLDZUQ3+1i//jIWZjB5mNbTwTVCmtAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"b6c18a35792376ca1b537974a3fbb2611876e15fd7f7ed7276e3dde75a33cdf6","last_reissued_at":"2026-06-30T02:17:20.138356Z","signature_status":"signed_v1","first_computed_at":"2026-06-30T02:17:20.138356Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2604.04385","source_version":5,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-30T02:17:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"s4LkoZvXYVgkLZhunwwMG3pJxyG8+mW/Mdic/YotnvGKJtxwibgYof/R9AmAeLyoFH+JRfzm/2tGQCq3fNTYDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-04T23:07:50.859972Z"},"content_sha256":"eb440c79472dad64f6e30a243e529b626956f2a76a79efb7365ba7fc9e9ab7a0","schema_version":"1.0","event_id":"sha256:eb440c79472dad64f6e30a243e529b626956f2a76a79efb7365ba7fc9e9ab7a0"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:W3AYUNLZEN3MUG2TPF2KH65SME","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities.","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Gregory N. Frank","submitted_at":"2026-04-06T03:20:37Z","abstract_excerpt":"We localize the policy routing mechanism in alignment-trained language models. An intermediate-layer attention gate reads detected content and triggers deeper amplifier heads that boost the signal toward refusal. In smaller models the gate and amplifier are single heads; at larger scale they become bands of heads across adjacent layers. The gate contributes under 1% of output DLA, yet interchange testing (p < 0.001) and knockout cascade confirm it is causally necessary. Interchange screening at n >= 120 detects the same motif in twelve models from six labs (2B to 72B), though specific heads di"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"The safety-trained capability is gated by routing, not removed; modulating the detection-layer signal continuously controls policy from hard refusal through evasion to factual answering, and any encoding that defeats detection-layer pattern matching bypasses the policy regardless of whether deeper layers reconstruct the content.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That interchange interventions and knockout cascades isolate the causal contribution of the identified gate and amplifier heads without substantial side effects on other circuits or on the model's general capability.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Alignment policy in language models is implemented as an early-commitment routing circuit of detection gates and amplifier heads that can be localized, scaled, and directly controlled without removing the underlying capability.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8a75a0afff28e65dfff5b83fa631dfa17be1766a2792cf7c54a9a579a85485a0"},"source":{"id":"2604.04385","kind":"arxiv","version":5},"verdict":{"id":"b218384b-8d03-456e-b3a9-fb3b2b359727","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-10T20:03:55.297389Z","strongest_claim":"The safety-trained capability is gated by routing, not removed; modulating the detection-layer signal continuously controls policy from hard refusal through evasion to factual answering, and any encoding that defeats detection-layer pattern matching bypasses the policy regardless of whether deeper layers reconstruct the content.","one_line_summary":"Alignment policy in language models is implemented as an early-commitment routing circuit of detection gates and amplifier heads that can be localized, scaled, and directly controlled without removing the underlying capability.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That interchange interventions and knockout cascades isolate the causal contribution of the identified gate and amplifier heads without substantial side effects on other circuits or on the model's general capability.","pith_extraction_headline":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.04385/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":"b218384b-8d03-456e-b3a9-fb3b2b359727"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-06-30T02:17:20Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"d0F2iJug5qKc8zAuQyId7WRHO5utXDHpLIAv3lh0LcZbQhYMYUPut4a3D+oOLRcBAOGRRmU7R21iG/UBY+uoDQ==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-07-04T23:07:50.860446Z"},"content_sha256":"20cc66e09e3a8cab058779d50150d70a5be76fd84236bbaa2e6e4b5b90086945","schema_version":"1.0","event_id":"sha256:20cc66e09e3a8cab058779d50150d70a5be76fd84236bbaa2e6e4b5b90086945"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/W3AYUNLZEN3MUG2TPF2KH65SME/bundle.json","state_url":"https://pith.science/pith/W3AYUNLZEN3MUG2TPF2KH65SME/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/W3AYUNLZEN3MUG2TPF2KH65SME/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":"2026-07-04T23:07:50Z","links":{"resolver":"https://pith.science/pith/W3AYUNLZEN3MUG2TPF2KH65SME","bundle":"https://pith.science/pith/W3AYUNLZEN3MUG2TPF2KH65SME/bundle.json","state":"https://pith.science/pith/W3AYUNLZEN3MUG2TPF2KH65SME/state.json","well_known_bundle":"https://pith.science/.well-known/pith/W3AYUNLZEN3MUG2TPF2KH65SME/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:W3AYUNLZEN3MUG2TPF2KH65SME","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":"f522f2b99c1fb0ffb7e53ab70c8fefe8c3473976e4c4ffa1703bcdf5cbd639ec","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-06T03:20:37Z","title_canon_sha256":"2535a2bd1b60beef235b319f09b2c08e0dc47b227dca504ee3bdfb81f47b9caf"},"schema_version":"1.0","source":{"id":"2604.04385","kind":"arxiv","version":5}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2604.04385","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"arxiv_version","alias_value":"2604.04385v5","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2604.04385","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_12","alias_value":"W3AYUNLZEN3M","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_16","alias_value":"W3AYUNLZEN3MUG2T","created_at":"2026-06-30T02:17:20Z"},{"alias_kind":"pith_short_8","alias_value":"W3AYUNLZ","created_at":"2026-06-30T02:17:20Z"}],"graph_snapshots":[{"event_id":"sha256:20cc66e09e3a8cab058779d50150d70a5be76fd84236bbaa2e6e4b5b90086945","target":"graph","created_at":"2026-06-30T02:17:20Z","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":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"The safety-trained capability is gated by routing, not removed; modulating the detection-layer signal continuously controls policy from hard refusal through evasion to factual answering, and any encoding that defeats detection-layer pattern matching bypasses the policy regardless of whether deeper layers reconstruct the content."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That interchange interventions and knockout cascades isolate the causal contribution of the identified gate and amplifier heads without substantial side effects on other circuits or on the model's general capability."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Alignment policy in language models is implemented as an early-commitment routing circuit of detection gates and amplifier heads that can be localized, scaled, and directly controlled without removing the underlying capability."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities."}],"snapshot_sha256":"8a75a0afff28e65dfff5b83fa631dfa17be1766a2792cf7c54a9a579a85485a0"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[],"endpoint":"/pith/2604.04385/integrity.json","findings":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"We localize the policy routing mechanism in alignment-trained language models. An intermediate-layer attention gate reads detected content and triggers deeper amplifier heads that boost the signal toward refusal. In smaller models the gate and amplifier are single heads; at larger scale they become bands of heads across adjacent layers. The gate contributes under 1% of output DLA, yet interchange testing (p < 0.001) and knockout cascade confirm it is causally necessary. Interchange screening at n >= 120 detects the same motif in twelve models from six labs (2B to 72B), though specific heads di","authors_text":"Gregory N. Frank","cross_cats":["cs.AI","cs.LG"],"headline":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities.","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-06T03:20:37Z","title":"How Alignment Routes: Localizing, Scaling, and Controlling Policy Circuits in Language Models"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2604.04385","kind":"arxiv","version":5},"verdict":{"created_at":"2026-05-10T20:03:55.297389Z","id":"b218384b-8d03-456e-b3a9-fb3b2b359727","model_set":{"reader":"grok-4.3"},"one_line_summary":"Alignment policy in language models is implemented as an early-commitment routing circuit of detection gates and amplifier heads that can be localized, scaled, and directly controlled without removing the underlying capability.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"Alignment in language models routes safety policies through early attention gates rather than erasing unsafe capabilities.","strongest_claim":"The safety-trained capability is gated by routing, not removed; modulating the detection-layer signal continuously controls policy from hard refusal through evasion to factual answering, and any encoding that defeats detection-layer pattern matching bypasses the policy regardless of whether deeper layers reconstruct the content.","weakest_assumption":"That interchange interventions and knockout cascades isolate the causal contribution of the identified gate and amplifier heads without substantial side effects on other circuits or on the model's general capability."}},"verdict_id":"b218384b-8d03-456e-b3a9-fb3b2b359727"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:eb440c79472dad64f6e30a243e529b626956f2a76a79efb7365ba7fc9e9ab7a0","target":"record","created_at":"2026-06-30T02:17:20Z","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":"f522f2b99c1fb0ffb7e53ab70c8fefe8c3473976e4c4ffa1703bcdf5cbd639ec","cross_cats_sorted":["cs.AI","cs.LG"],"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2026-04-06T03:20:37Z","title_canon_sha256":"2535a2bd1b60beef235b319f09b2c08e0dc47b227dca504ee3bdfb81f47b9caf"},"schema_version":"1.0","source":{"id":"2604.04385","kind":"arxiv","version":5}},"canonical_sha256":"b6c18a35792376ca1b537974a3fbb2611876e15fd7f7ed7276e3dde75a33cdf6","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"b6c18a35792376ca1b537974a3fbb2611876e15fd7f7ed7276e3dde75a33cdf6","first_computed_at":"2026-06-30T02:17:20.138356Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-06-30T02:17:20.138356Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"nYRdF2mIY5V/2hQo8c+gZyRvrMxbR+V4aLkPY6LkP31HQsxmyFBOpzSfLDZUQ3+1i//jIWZjB5mNbTwTVCmtAg==","signature_status":"signed_v1","signed_at":"2026-06-30T02:17:20.139021Z","signed_message":"canonical_sha256_bytes"},"source_id":"2604.04385","source_kind":"arxiv","source_version":5}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:eb440c79472dad64f6e30a243e529b626956f2a76a79efb7365ba7fc9e9ab7a0","sha256:20cc66e09e3a8cab058779d50150d70a5be76fd84236bbaa2e6e4b5b90086945"],"state_sha256":"36d5785bde583a630acf0593452bc64b22ab8f7663411a3c714f7817849c962c"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"JqHOeiF9cQzSVGknk5XEKp0fH4SsbKATnY3j2GrOsRepXLJabN22m2IP98srvt1tvwOLaczqbE50chxLnfr3CA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-07-04T23:07:50.863124Z","bundle_sha256":"26bd743ae050dda82cbf99c408ea1f9840a2b9da3a89e7404dd6692fa4ee41fc"}}