{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:QF6QSDU4TLPHMHTNREPBWVW5HS","short_pith_number":"pith:QF6QSDU4","schema_version":"1.0","canonical_sha256":"817d090e9c9ade761e6d891e1b56dd3cb783a5b2270024ca2aaef37ef690c803","source":{"kind":"arxiv","id":"2403.05767","version":1},"attestation_state":"computed","paper":{"title":"Extending Activation Steering to Broad Skills and Multiple Behaviours","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CY"],"primary_cat":"cs.LG","authors_text":"Massimo Poesio, Nandi Schoots, Teun van der Weij","submitted_at":"2024-03-09T02:30:04Z","abstract_excerpt":"Current large language models have dangerous capabilities, which are likely to become more problematic in the future. Activation steering techniques can be used to reduce risks from these capabilities. In this paper, we investigate the efficacy of activation steering for broad skills and multiple behaviours. First, by comparing the effects of reducing performance on general coding ability and Python-specific ability, we find that steering broader skills is competitive to steering narrower skills. Second, we steer models to become more or less myopic and wealth-seeking, among other behaviours. "},"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":"2403.05767","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-03-09T02:30:04Z","cross_cats_sorted":["cs.AI","cs.CL","cs.CY"],"title_canon_sha256":"0448e4ae098b741107dbafd5a2784c9095a8bc0c5b8152ed212f5f7d4c7713e0","abstract_canon_sha256":"dea61c4a10ddf17506486b9576701df6af17a02c858dc534a7d9ff37c8439cd6"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:54:13.068646Z","signature_b64":"FI+znSOLFB+lLXGsv7vw9/AOynU1yb+pERmwJNFzAyKNnkD95/bX8z/cwKv+1RbfNGBQ0eE8Ks1vbmwVSIP+Dw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"817d090e9c9ade761e6d891e1b56dd3cb783a5b2270024ca2aaef37ef690c803","last_reissued_at":"2026-07-05T07:54:13.068213Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:54:13.068213Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Extending Activation Steering to Broad Skills and Multiple Behaviours","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL","cs.CY"],"primary_cat":"cs.LG","authors_text":"Massimo Poesio, Nandi Schoots, Teun van der Weij","submitted_at":"2024-03-09T02:30:04Z","abstract_excerpt":"Current large language models have dangerous capabilities, which are likely to become more problematic in the future. Activation steering techniques can be used to reduce risks from these capabilities. In this paper, we investigate the efficacy of activation steering for broad skills and multiple behaviours. First, by comparing the effects of reducing performance on general coding ability and Python-specific ability, we find that steering broader skills is competitive to steering narrower skills. Second, we steer models to become more or less myopic and wealth-seeking, among other behaviours. "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.05767","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/2403.05767/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":"2403.05767","created_at":"2026-07-05T07:54:13.068270+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.05767v1","created_at":"2026-07-05T07:54:13.068270+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.05767","created_at":"2026-07-05T07:54:13.068270+00:00"},{"alias_kind":"pith_short_12","alias_value":"QF6QSDU4TLPH","created_at":"2026-07-05T07:54:13.068270+00:00"},{"alias_kind":"pith_short_16","alias_value":"QF6QSDU4TLPHMHTN","created_at":"2026-07-05T07:54:13.068270+00:00"},{"alias_kind":"pith_short_8","alias_value":"QF6QSDU4","created_at":"2026-07-05T07:54:13.068270+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":8,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.12234","citing_title":"On The Effectiveness-Fluency Trade-Off In LLM Conditioning: A Systematic Study","ref_index":106,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06214","citing_title":"Towards the Readability of LLM-Generated Codes through Multitask Representation Engineering","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17231","citing_title":"FishBack: Pullback Fisher Geometry for Optimal Activation Steering in Transformers","ref_index":46,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10664","citing_title":"Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.10664","citing_title":"Prompt-Activation Duality: Improving Activation Steering via Attention-Level Interventions","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24693","citing_title":"Contextual Linear Activation Steering of Language Models","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2604.06377","citing_title":"The Master Key Hypothesis: Unlocking Cross-Model Capability Transfer via Linear Subspace Alignment","ref_index":67,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03907","citing_title":"Steer Like the LLM: Activation Steering that Mimics Prompting","ref_index":24,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS","json":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS.json","graph_json":"https://pith.science/api/pith-number/QF6QSDU4TLPHMHTNREPBWVW5HS/graph.json","events_json":"https://pith.science/api/pith-number/QF6QSDU4TLPHMHTNREPBWVW5HS/events.json","paper":"https://pith.science/paper/QF6QSDU4"},"agent_actions":{"view_html":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS","download_json":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS.json","view_paper":"https://pith.science/paper/QF6QSDU4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.05767&json=true","fetch_graph":"https://pith.science/api/pith-number/QF6QSDU4TLPHMHTNREPBWVW5HS/graph.json","fetch_events":"https://pith.science/api/pith-number/QF6QSDU4TLPHMHTNREPBWVW5HS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS/action/storage_attestation","attest_author":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS/action/author_attestation","sign_citation":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS/action/citation_signature","submit_replication":"https://pith.science/pith/QF6QSDU4TLPHMHTNREPBWVW5HS/action/replication_record"}},"created_at":"2026-07-05T07:54:13.068270+00:00","updated_at":"2026-07-05T07:54:13.068270+00:00"}