{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:VMUJYAMUBWAIONZ3M6A6YCMCRN","short_pith_number":"pith:VMUJYAMU","schema_version":"1.0","canonical_sha256":"ab289c01940d8087373b6781ec09828b6e87d8f7b556287c53a464acb8bffd25","source":{"kind":"arxiv","id":"2607.13075","version":1},"attestation_state":"computed","paper":{"title":"The Entanglement Wall: Activation-Space Probes as Risk Detectors, Not Context Adjudicators","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Dominik Schwarz","submitted_at":"2026-07-12T20:37:19Z","abstract_excerpt":"Context can change whether a request is harmful without changing its topic or surface form. We ask whether residual-stream probes distinguish harmful requests from surface-matched benign controls at a useful operating point. Across three 7-8B model families, an activation sensor blocks 95.5-97.7 percent of judge-classified compliant attacks in a taxonomy-selected set. It also blocks 59.6-68.4 percent of XSTest prompts. A fully disjoint audit reconstructs near-ceiling source-contrast AUROC (0.996-0.999), but fixed transfer to matched pairs is weaker: 0.656-0.819 on the guard-selected Twin-n70 s"},"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":"2607.13075","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CR","submitted_at":"2026-07-12T20:37:19Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"6729766bf0b74202e8167b479ec4396b19dfa4bf692d4ae9956ce2590cfdda17","abstract_canon_sha256":"a3d5415ec1f596dd70774004da02a520df5f277db59d545169ca8ea417d118e3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-16T00:21:54.591372Z","signature_b64":"mBEOjqBBD3VXe1k9bOVUrsylnCWioiPKceh7BerfkU+It12UmbdG9IWh6lwF5+Sj2/63h3Lx0P0Z/+/1DF6JCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ab289c01940d8087373b6781ec09828b6e87d8f7b556287c53a464acb8bffd25","last_reissued_at":"2026-07-16T00:21:54.590433Z","signature_status":"signed_v1","first_computed_at":"2026-07-16T00:21:54.590433Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Entanglement Wall: Activation-Space Probes as Risk Detectors, Not Context Adjudicators","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CR","authors_text":"Dominik Schwarz","submitted_at":"2026-07-12T20:37:19Z","abstract_excerpt":"Context can change whether a request is harmful without changing its topic or surface form. We ask whether residual-stream probes distinguish harmful requests from surface-matched benign controls at a useful operating point. Across three 7-8B model families, an activation sensor blocks 95.5-97.7 percent of judge-classified compliant attacks in a taxonomy-selected set. It also blocks 59.6-68.4 percent of XSTest prompts. A fully disjoint audit reconstructs near-ceiling source-contrast AUROC (0.996-0.999), but fixed transfer to matched pairs is weaker: 0.656-0.819 on the guard-selected Twin-n70 s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.13075","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/2607.13075/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":"2607.13075","created_at":"2026-07-16T00:21:54.590882+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.13075v1","created_at":"2026-07-16T00:21:54.590882+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.13075","created_at":"2026-07-16T00:21:54.590882+00:00"},{"alias_kind":"pith_short_12","alias_value":"VMUJYAMUBWAI","created_at":"2026-07-16T00:21:54.590882+00:00"},{"alias_kind":"pith_short_16","alias_value":"VMUJYAMUBWAIONZ3","created_at":"2026-07-16T00:21:54.590882+00:00"},{"alias_kind":"pith_short_8","alias_value":"VMUJYAMU","created_at":"2026-07-16T00:21:54.590882+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/VMUJYAMUBWAIONZ3M6A6YCMCRN","json":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN.json","graph_json":"https://pith.science/api/pith-number/VMUJYAMUBWAIONZ3M6A6YCMCRN/graph.json","events_json":"https://pith.science/api/pith-number/VMUJYAMUBWAIONZ3M6A6YCMCRN/events.json","paper":"https://pith.science/paper/VMUJYAMU"},"agent_actions":{"view_html":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN","download_json":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN.json","view_paper":"https://pith.science/paper/VMUJYAMU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.13075&json=true","fetch_graph":"https://pith.science/api/pith-number/VMUJYAMUBWAIONZ3M6A6YCMCRN/graph.json","fetch_events":"https://pith.science/api/pith-number/VMUJYAMUBWAIONZ3M6A6YCMCRN/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN/action/timestamp_anchor","attest_storage":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN/action/storage_attestation","attest_author":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN/action/author_attestation","sign_citation":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN/action/citation_signature","submit_replication":"https://pith.science/pith/VMUJYAMUBWAIONZ3M6A6YCMCRN/action/replication_record"}},"created_at":"2026-07-16T00:21:54.590882+00:00","updated_at":"2026-07-16T00:21:54.590882+00:00"}