{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:3LPRHOJ45FOAAMUJ3SZAVU7E2W","short_pith_number":"pith:3LPRHOJ4","schema_version":"1.0","canonical_sha256":"dadf13b93ce95c003289dcb20ad3e4d5a51398b238fc2554cc3342f516bf3131","source":{"kind":"arxiv","id":"2606.08454","version":1},"attestation_state":"computed","paper":{"title":"Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Thai Le, Tuc Nguyen","submitted_at":"2026-06-07T05:01:25Z","abstract_excerpt":"Activation steering provides a lightweight inference-time mechanism for controlling large language models (LLMs) by modifying their internal activation vectors toward desired behaviors. Most existing methods compute a fixed steering direction in the original activation space, typically from pairs of contrastive examples using mean differences, linear probes, or arbitrary separability criteria. While effective to a certain extent, these methods treat behavioral control as a global, linear, additive offset: the same direction is applied across inputs, and behaviors are linearly separable. This c"},"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":"2606.08454","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2026-06-07T05:01:25Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"020d831057bf5872d6c454dbf927c08ad84ab20fd04fe30520e771c7775e381d","abstract_canon_sha256":"f75c7d7670305ff77afadcf452bb05c45b65d9b9c9e7b6a091f05839febb158d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-09T01:05:37.081858Z","signature_b64":"KL8NITfOwxTAeusUErfNrzIbRQEsYIg1n7u0s0T3K8hbj7sWqzBFNQ6ZnUdcfbiZtcwaFwyuY8s2HsuSlRejBw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"dadf13b93ce95c003289dcb20ad3e4d5a51398b238fc2554cc3342f516bf3131","last_reissued_at":"2026-06-09T01:05:37.081482Z","signature_status":"signed_v1","first_computed_at":"2026-06-09T01:05:37.081482Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Beyond Linear Activation Steering: Invertible Latent Transformations for Controlling LLM Behavior","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.LG","authors_text":"Thai Le, Tuc Nguyen","submitted_at":"2026-06-07T05:01:25Z","abstract_excerpt":"Activation steering provides a lightweight inference-time mechanism for controlling large language models (LLMs) by modifying their internal activation vectors toward desired behaviors. Most existing methods compute a fixed steering direction in the original activation space, typically from pairs of contrastive examples using mean differences, linear probes, or arbitrary separability criteria. While effective to a certain extent, these methods treat behavioral control as a global, linear, additive offset: the same direction is applied across inputs, and behaviors are linearly separable. This c"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.08454","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/2606.08454/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":"2606.08454","created_at":"2026-06-09T01:05:37.081543+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.08454v1","created_at":"2026-06-09T01:05:37.081543+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.08454","created_at":"2026-06-09T01:05:37.081543+00:00"},{"alias_kind":"pith_short_12","alias_value":"3LPRHOJ45FOA","created_at":"2026-06-09T01:05:37.081543+00:00"},{"alias_kind":"pith_short_16","alias_value":"3LPRHOJ45FOAAMUJ","created_at":"2026-06-09T01:05:37.081543+00:00"},{"alias_kind":"pith_short_8","alias_value":"3LPRHOJ4","created_at":"2026-06-09T01:05:37.081543+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/3LPRHOJ45FOAAMUJ3SZAVU7E2W","json":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W.json","graph_json":"https://pith.science/api/pith-number/3LPRHOJ45FOAAMUJ3SZAVU7E2W/graph.json","events_json":"https://pith.science/api/pith-number/3LPRHOJ45FOAAMUJ3SZAVU7E2W/events.json","paper":"https://pith.science/paper/3LPRHOJ4"},"agent_actions":{"view_html":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W","download_json":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W.json","view_paper":"https://pith.science/paper/3LPRHOJ4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.08454&json=true","fetch_graph":"https://pith.science/api/pith-number/3LPRHOJ45FOAAMUJ3SZAVU7E2W/graph.json","fetch_events":"https://pith.science/api/pith-number/3LPRHOJ45FOAAMUJ3SZAVU7E2W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W/action/storage_attestation","attest_author":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W/action/author_attestation","sign_citation":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W/action/citation_signature","submit_replication":"https://pith.science/pith/3LPRHOJ45FOAAMUJ3SZAVU7E2W/action/replication_record"}},"created_at":"2026-06-09T01:05:37.081543+00:00","updated_at":"2026-06-09T01:05:37.081543+00:00"}