{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:PJYQ6GPBZJDAIQBJS5RBLVU276","short_pith_number":"pith:PJYQ6GPB","schema_version":"1.0","canonical_sha256":"7a710f19e1ca46044029976215d69affbccb0d06ebef06821ab37dc8e3c1ebbc","source":{"kind":"arxiv","id":"2605.28664","version":1},"attestation_state":"computed","paper":{"title":"Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Anna Rumshisky, Leman Akoglu, Tootiya Giyahchi, Veena Padmanabhan, Vijeta Deshpande","submitted_at":"2026-05-27T15:59:45Z","abstract_excerpt":"Safety detection models require examples of HHH (Helpful, Harmless, Honest)-violating outputs for robust generalization, however such examples are scarce. Activation Steering (AS) has emerged as a data-efficient method for generating target-concept-aligned responses. We investigate whether AS can generate high-quality training datasets for downstream classifiers, a question that remains untested. We present a two-fold study with intrinsic and extrinsic evaluation across $4$ concepts $\\times\\,2$ models $\\times\\,4$ steering methods. Intrinsically, beyond the field-standard rubric of steering suc"},"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":"2605.28664","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-27T15:59:45Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"e78a12aef9399632df4d159c8bcc8e1d358782af3fea44b4c153913690d77d24","abstract_canon_sha256":"a64ec8fe99eb87559f967005658d6fec4593f3fedff86588bdb086cfb926f6c5"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-28T02:04:59.252672Z","signature_b64":"4KcRfICEJp9etHmNMOWT3thucW+th32YEZ7XD8WPD8EiDVKBx5TSIHElzAjy/ckvjRNHUJkLhsIk9OvUuK0gDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7a710f19e1ca46044029976215d69affbccb0d06ebef06821ab37dc8e3c1ebbc","last_reissued_at":"2026-05-28T02:04:59.252223Z","signature_status":"signed_v1","first_computed_at":"2026-05-28T02:04:59.252223Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Anna Rumshisky, Leman Akoglu, Tootiya Giyahchi, Veena Padmanabhan, Vijeta Deshpande","submitted_at":"2026-05-27T15:59:45Z","abstract_excerpt":"Safety detection models require examples of HHH (Helpful, Harmless, Honest)-violating outputs for robust generalization, however such examples are scarce. Activation Steering (AS) has emerged as a data-efficient method for generating target-concept-aligned responses. We investigate whether AS can generate high-quality training datasets for downstream classifiers, a question that remains untested. We present a two-fold study with intrinsic and extrinsic evaluation across $4$ concepts $\\times\\,2$ models $\\times\\,4$ steering methods. Intrinsically, beyond the field-standard rubric of steering suc"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.28664","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/2605.28664/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":"2605.28664","created_at":"2026-05-28T02:04:59.252281+00:00"},{"alias_kind":"arxiv_version","alias_value":"2605.28664v1","created_at":"2026-05-28T02:04:59.252281+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.28664","created_at":"2026-05-28T02:04:59.252281+00:00"},{"alias_kind":"pith_short_12","alias_value":"PJYQ6GPBZJDA","created_at":"2026-05-28T02:04:59.252281+00:00"},{"alias_kind":"pith_short_16","alias_value":"PJYQ6GPBZJDAIQBJ","created_at":"2026-05-28T02:04:59.252281+00:00"},{"alias_kind":"pith_short_8","alias_value":"PJYQ6GPB","created_at":"2026-05-28T02:04:59.252281+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/PJYQ6GPBZJDAIQBJS5RBLVU276","json":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276.json","graph_json":"https://pith.science/api/pith-number/PJYQ6GPBZJDAIQBJS5RBLVU276/graph.json","events_json":"https://pith.science/api/pith-number/PJYQ6GPBZJDAIQBJS5RBLVU276/events.json","paper":"https://pith.science/paper/PJYQ6GPB"},"agent_actions":{"view_html":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276","download_json":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276.json","view_paper":"https://pith.science/paper/PJYQ6GPB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2605.28664&json=true","fetch_graph":"https://pith.science/api/pith-number/PJYQ6GPBZJDAIQBJS5RBLVU276/graph.json","fetch_events":"https://pith.science/api/pith-number/PJYQ6GPBZJDAIQBJS5RBLVU276/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276/action/storage_attestation","attest_author":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276/action/author_attestation","sign_citation":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276/action/citation_signature","submit_replication":"https://pith.science/pith/PJYQ6GPBZJDAIQBJS5RBLVU276/action/replication_record"}},"created_at":"2026-05-28T02:04:59.252281+00:00","updated_at":"2026-05-28T02:04:59.252281+00:00"}