{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:FSLJZQXSN4NPKJVPDOTMPCPN6T","short_pith_number":"pith:FSLJZQXS","schema_version":"1.0","canonical_sha256":"2c969cc2f26f1af526af1ba6c789edf4eab82f08aefbc91dc453a06ad9522a07","source":{"kind":"arxiv","id":"2602.06941","version":2},"attestation_state":"computed","paper":{"title":"Endogenous Resistance to Activation Steering in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alex McKenzie, Diogo de Lucena, Judd Rosenblatt, Keenan Pepper, Martin Leitgab, Michael S. A. Graziano, Mike Vaiana, Murat Cubuktepe, Stijn Servaes","submitted_at":"2026-02-06T18:41:12Z","abstract_excerpt":"Large language models can recover mid-generation from task-misaligned activation steering, producing explicit verbal restarts (e.g., ``wait, that's not right'') and continuing on-topic even while the steering perturbation remains active. We term this Endogenous Steering Resistance (ESR). Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B exhibits explicit ESR at \\llamaseventyEsrRate\\%, with smaller models from the Llama-3 and Gemma-2 families showing the explicit form less frequently. Two controls dissociate ESR into a detection event and a sustained-"},"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":"2602.06941","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-02-06T18:41:12Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"0e0340754280d833c21176fafedad039417ef09f42f1e56213519c0373759272","abstract_canon_sha256":"283d474a2aa49a59d89c3a6bbd4b1b27738a66a9891f67913bab0f57540e02f7"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-08T01:03:58.019788Z","signature_b64":"mQ6uJ3glpV9sbxy+Rao7a1wq4aHtAdrJP8bq4z130h3PGVR9u1HRXjen5c9zEiZPBytjBa4VW4++ZRajER/uDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2c969cc2f26f1af526af1ba6c789edf4eab82f08aefbc91dc453a06ad9522a07","last_reissued_at":"2026-06-08T01:03:58.018854Z","signature_status":"signed_v1","first_computed_at":"2026-06-08T01:03:58.018854Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Endogenous Resistance to Activation Steering in Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Alex McKenzie, Diogo de Lucena, Judd Rosenblatt, Keenan Pepper, Martin Leitgab, Michael S. A. Graziano, Mike Vaiana, Murat Cubuktepe, Stijn Servaes","submitted_at":"2026-02-06T18:41:12Z","abstract_excerpt":"Large language models can recover mid-generation from task-misaligned activation steering, producing explicit verbal restarts (e.g., ``wait, that's not right'') and continuing on-topic even while the steering perturbation remains active. We term this Endogenous Steering Resistance (ESR). Using sparse autoencoder (SAE) latents to steer model activations, we find that Llama-3.3-70B exhibits explicit ESR at \\llamaseventyEsrRate\\%, with smaller models from the Llama-3 and Gemma-2 families showing the explicit form less frequently. Two controls dissociate ESR into a detection event and a sustained-"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2602.06941","kind":"arxiv","version":2},"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/2602.06941/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":"2602.06941","created_at":"2026-06-08T01:03:58.018977+00:00"},{"alias_kind":"arxiv_version","alias_value":"2602.06941v2","created_at":"2026-06-08T01:03:58.018977+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2602.06941","created_at":"2026-06-08T01:03:58.018977+00:00"},{"alias_kind":"pith_short_12","alias_value":"FSLJZQXSN4NP","created_at":"2026-06-08T01:03:58.018977+00:00"},{"alias_kind":"pith_short_16","alias_value":"FSLJZQXSN4NPKJVP","created_at":"2026-06-08T01:03:58.018977+00:00"},{"alias_kind":"pith_short_8","alias_value":"FSLJZQXS","created_at":"2026-06-08T01:03:58.018977+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2605.05715","citing_title":"Decodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes","ref_index":45,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T","json":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T.json","graph_json":"https://pith.science/api/pith-number/FSLJZQXSN4NPKJVPDOTMPCPN6T/graph.json","events_json":"https://pith.science/api/pith-number/FSLJZQXSN4NPKJVPDOTMPCPN6T/events.json","paper":"https://pith.science/paper/FSLJZQXS"},"agent_actions":{"view_html":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T","download_json":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T.json","view_paper":"https://pith.science/paper/FSLJZQXS","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2602.06941&json=true","fetch_graph":"https://pith.science/api/pith-number/FSLJZQXSN4NPKJVPDOTMPCPN6T/graph.json","fetch_events":"https://pith.science/api/pith-number/FSLJZQXSN4NPKJVPDOTMPCPN6T/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T/action/storage_attestation","attest_author":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T/action/author_attestation","sign_citation":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T/action/citation_signature","submit_replication":"https://pith.science/pith/FSLJZQXSN4NPKJVPDOTMPCPN6T/action/replication_record"}},"created_at":"2026-06-08T01:03:58.018977+00:00","updated_at":"2026-06-08T01:03:58.018977+00:00"}