{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:JQQDZSXII2CLETKTYUI2O3IGMD","short_pith_number":"pith:JQQDZSXI","schema_version":"1.0","canonical_sha256":"4c203ccae84684b24d53c511a76d0660e7563f0446a99d0307826af9f42540cc","source":{"kind":"arxiv","id":"2506.03292","version":1},"attestation_state":"computed","paper":{"title":"HyperSteer: Activation Steering at Scale with Hypernetworks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Christopher Potts, Jiuding Sun, Michael Sklar, Sidharth Baskaran, Zhengxuan Wu","submitted_at":"2025-06-03T18:32:01Z","abstract_excerpt":"Steering language models (LMs) by modifying internal activations is a popular approach for controlling text generation. Unsupervised dictionary learning methods, e.g., sparse autoencoders, can be scaled to produce many steering vectors, but lack guarantees on the individual efficacy of each vector and control over the coverage of relevant steering tasks. In contrast, supervised methods for constructing steering vectors are targeted and effective, but require more data collection and training for each additional steering vector produced. In this work, we introduce HyperSteer, a family of hypern"},"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":"2506.03292","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-03T18:32:01Z","cross_cats_sorted":["cs.AI","cs.LG"],"title_canon_sha256":"ce964d7e11cae485d2660b34061f9719578e854c398c88ec3a91204428bc2636","abstract_canon_sha256":"fbe0bb0913667374e0dfcffdf85fc64651d05e4f17295ac1a8ca4a68e4227b7d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:15:24.299263Z","signature_b64":"xooAaIx3321SdOuTsLl1+JDOZdPQrPiIBV33A/oDTFbTFgD82EP8VzLszrpP860aFgkNNNz846XfehL/rkddDg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4c203ccae84684b24d53c511a76d0660e7563f0446a99d0307826af9f42540cc","last_reissued_at":"2026-07-05T11:15:24.298773Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:15:24.298773Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"HyperSteer: Activation Steering at Scale with Hypernetworks","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.LG"],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Christopher Potts, Jiuding Sun, Michael Sklar, Sidharth Baskaran, Zhengxuan Wu","submitted_at":"2025-06-03T18:32:01Z","abstract_excerpt":"Steering language models (LMs) by modifying internal activations is a popular approach for controlling text generation. Unsupervised dictionary learning methods, e.g., sparse autoencoders, can be scaled to produce many steering vectors, but lack guarantees on the individual efficacy of each vector and control over the coverage of relevant steering tasks. In contrast, supervised methods for constructing steering vectors are targeted and effective, but require more data collection and training for each additional steering vector produced. In this work, we introduce HyperSteer, a family of hypern"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.03292","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/2506.03292/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":"2506.03292","created_at":"2026-07-05T11:15:24.298836+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.03292v1","created_at":"2026-07-05T11:15:24.298836+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.03292","created_at":"2026-07-05T11:15:24.298836+00:00"},{"alias_kind":"pith_short_12","alias_value":"JQQDZSXII2CL","created_at":"2026-07-05T11:15:24.298836+00:00"},{"alias_kind":"pith_short_16","alias_value":"JQQDZSXII2CLETKT","created_at":"2026-07-05T11:15:24.298836+00:00"},{"alias_kind":"pith_short_8","alias_value":"JQQDZSXI","created_at":"2026-07-05T11:15:24.298836+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24154","citing_title":"Palette: A Modular, Controllable, and Efficient Framework for On-demand Authorized Safety Alignment Relaxation in LLMs","ref_index":64,"is_internal_anchor":false},{"citing_arxiv_id":"2605.28664","citing_title":"Activation Steering for Synthetic Data Generation: The Role of Diversity in Downstream Safety Detection","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16362","citing_title":"When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16362","citing_title":"When Is Rank-1 Steering Cheap? Geometry, Granularity, and Budgeted Search","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.08254","citing_title":"HyperTransport: Amortized Conditioning of T2I Generative Models","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2604.08524","citing_title":"What Drives Representation Steering? A Mechanistic Case Study on Steering Refusal","ref_index":33,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03907","citing_title":"Steer Like the LLM: Activation Steering that Mimics Prompting","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD","json":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD.json","graph_json":"https://pith.science/api/pith-number/JQQDZSXII2CLETKTYUI2O3IGMD/graph.json","events_json":"https://pith.science/api/pith-number/JQQDZSXII2CLETKTYUI2O3IGMD/events.json","paper":"https://pith.science/paper/JQQDZSXI"},"agent_actions":{"view_html":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD","download_json":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD.json","view_paper":"https://pith.science/paper/JQQDZSXI","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.03292&json=true","fetch_graph":"https://pith.science/api/pith-number/JQQDZSXII2CLETKTYUI2O3IGMD/graph.json","fetch_events":"https://pith.science/api/pith-number/JQQDZSXII2CLETKTYUI2O3IGMD/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD/action/storage_attestation","attest_author":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD/action/author_attestation","sign_citation":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD/action/citation_signature","submit_replication":"https://pith.science/pith/JQQDZSXII2CLETKTYUI2O3IGMD/action/replication_record"}},"created_at":"2026-07-05T11:15:24.298836+00:00","updated_at":"2026-07-05T11:15:24.298836+00:00"}