{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:RB3ZOVRBXVJPC25FDBIILDCJTA","short_pith_number":"pith:RB3ZOVRB","schema_version":"1.0","canonical_sha256":"8877975621bd52f16ba51850858c499802f40fd1b8a2a64c58b628b6c851c1e8","source":{"kind":"arxiv","id":"2307.00787","version":1},"attestation_state":"computed","paper":{"title":"Evaluating Shutdown Avoidance of Language Models in Textual Scenarios","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Leon Lang, Simon Lermen, Teun van der Weij","submitted_at":"2023-07-03T07:05:59Z","abstract_excerpt":"Recently, there has been an increase in interest in evaluating large language models for emergent and dangerous capabilities. Importantly, agents could reason that in some scenarios their goal is better achieved if they are not turned off, which can lead to undesirable behaviors. In this paper, we investigate the potential of using toy textual scenarios to evaluate instrumental reasoning and shutdown avoidance in language models such as GPT-4 and Claude. Furthermore, we explore whether shutdown avoidance is merely a result of simple pattern matching between the dataset and the prompt or if it "},"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":"2307.00787","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-07-03T07:05:59Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"e129a2e3eed80cdc2bc76c9b96d05fd077dda25108b2ddbe9915b63dd1fb19e1","abstract_canon_sha256":"b2a06f95e2aaa7c192c986715ff7faaba3d500d7af492a55ef1b99fc0ec0eb79"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T06:27:00.507754Z","signature_b64":"+EYbOTs0kcSjLfQBgT43leJci7mK5n1rBjSeMAZClL/GeM5pgO/acQioh3rFaJ6rhG9TIaIingMlCgX0n7AQDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8877975621bd52f16ba51850858c499802f40fd1b8a2a64c58b628b6c851c1e8","last_reissued_at":"2026-07-05T06:27:00.507275Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T06:27:00.507275Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evaluating Shutdown Avoidance of Language Models in Textual Scenarios","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Leon Lang, Simon Lermen, Teun van der Weij","submitted_at":"2023-07-03T07:05:59Z","abstract_excerpt":"Recently, there has been an increase in interest in evaluating large language models for emergent and dangerous capabilities. Importantly, agents could reason that in some scenarios their goal is better achieved if they are not turned off, which can lead to undesirable behaviors. In this paper, we investigate the potential of using toy textual scenarios to evaluate instrumental reasoning and shutdown avoidance in language models such as GPT-4 and Claude. Furthermore, we explore whether shutdown avoidance is merely a result of simple pattern matching between the dataset and the prompt or if it "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2307.00787","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/2307.00787/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":"2307.00787","created_at":"2026-07-05T06:27:00.507340+00:00"},{"alias_kind":"arxiv_version","alias_value":"2307.00787v1","created_at":"2026-07-05T06:27:00.507340+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2307.00787","created_at":"2026-07-05T06:27:00.507340+00:00"},{"alias_kind":"pith_short_12","alias_value":"RB3ZOVRBXVJP","created_at":"2026-07-05T06:27:00.507340+00:00"},{"alias_kind":"pith_short_16","alias_value":"RB3ZOVRBXVJPC25F","created_at":"2026-07-05T06:27:00.507340+00:00"},{"alias_kind":"pith_short_8","alias_value":"RB3ZOVRB","created_at":"2026-07-05T06:27:00.507340+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.09711","citing_title":"Proxy Reward Internalization and Mechanistic Exploitation: A Learned Precursor to Reward Hacking and Its Generalization","ref_index":245,"is_internal_anchor":false},{"citing_arxiv_id":"2604.13602","citing_title":"Reward Hacking in the Era of Large Models: Mechanisms, Emergent Misalignment, Challenges","ref_index":70,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA","json":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA.json","graph_json":"https://pith.science/api/pith-number/RB3ZOVRBXVJPC25FDBIILDCJTA/graph.json","events_json":"https://pith.science/api/pith-number/RB3ZOVRBXVJPC25FDBIILDCJTA/events.json","paper":"https://pith.science/paper/RB3ZOVRB"},"agent_actions":{"view_html":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA","download_json":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA.json","view_paper":"https://pith.science/paper/RB3ZOVRB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2307.00787&json=true","fetch_graph":"https://pith.science/api/pith-number/RB3ZOVRBXVJPC25FDBIILDCJTA/graph.json","fetch_events":"https://pith.science/api/pith-number/RB3ZOVRBXVJPC25FDBIILDCJTA/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA/action/storage_attestation","attest_author":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA/action/author_attestation","sign_citation":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA/action/citation_signature","submit_replication":"https://pith.science/pith/RB3ZOVRBXVJPC25FDBIILDCJTA/action/replication_record"}},"created_at":"2026-07-05T06:27:00.507340+00:00","updated_at":"2026-07-05T06:27:00.507340+00:00"}