{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KOTFVC4KHSF3KMDSLNGP4HO5AB","short_pith_number":"pith:KOTFVC4K","schema_version":"1.0","canonical_sha256":"53a65a8b8a3c8bb530725b4cfe1ddd00446dc3ddb0b08b7ff1e728f28e4e7fae","source":{"kind":"arxiv","id":"2505.02709","version":1},"attestation_state":"computed","paper":{"title":"Technical Report: Evaluating Goal Drift in Language Model Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Elizabeth Donoway, Henning Bartsch, Marius Hobbhahn, Rauno Arike","submitted_at":"2025-05-05T15:06:09Z","abstract_excerpt":"As language models (LMs) are increasingly deployed as autonomous agents, their robust adherence to human-assigned objectives becomes crucial for safe operation. When these agents operate independently for extended periods without human oversight, even initially well-specified goals may gradually shift. Detecting and measuring goal drift - an agent's tendency to deviate from its original objective over time - presents significant challenges, as goals can shift gradually, causing only subtle behavioral changes. This paper proposes a novel approach to analyzing goal drift in LM agents. In our exp"},"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":"2505.02709","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2025-05-05T15:06:09Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"14ddcd851347fcfc9329e2fe6fee8b88e92c222ed3d01e21199335507d0db1bc","abstract_canon_sha256":"51fbba4c9550f962be3fc7d9679d0d0e27d197bb2e23a5f9a5f03683fd70306b"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:58:41.658808Z","signature_b64":"3/H+mtF0SBqyeAo2PASIlSPDfrGgy2smL8VgPzlUoc7akvj6yKiamM3316NUI7huFMvPyb++X7iIfOX4n9FiCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"53a65a8b8a3c8bb530725b4cfe1ddd00446dc3ddb0b08b7ff1e728f28e4e7fae","last_reissued_at":"2026-07-05T10:58:41.658297Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:58:41.658297Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Technical Report: Evaluating Goal Drift in Language Model Agents","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.AI","authors_text":"Elizabeth Donoway, Henning Bartsch, Marius Hobbhahn, Rauno Arike","submitted_at":"2025-05-05T15:06:09Z","abstract_excerpt":"As language models (LMs) are increasingly deployed as autonomous agents, their robust adherence to human-assigned objectives becomes crucial for safe operation. When these agents operate independently for extended periods without human oversight, even initially well-specified goals may gradually shift. Detecting and measuring goal drift - an agent's tendency to deviate from its original objective over time - presents significant challenges, as goals can shift gradually, causing only subtle behavioral changes. This paper proposes a novel approach to analyzing goal drift in LM agents. In our exp"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.02709","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/2505.02709/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":"2505.02709","created_at":"2026-07-05T10:58:41.658369+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.02709v1","created_at":"2026-07-05T10:58:41.658369+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.02709","created_at":"2026-07-05T10:58:41.658369+00:00"},{"alias_kind":"pith_short_12","alias_value":"KOTFVC4KHSF3","created_at":"2026-07-05T10:58:41.658369+00:00"},{"alias_kind":"pith_short_16","alias_value":"KOTFVC4KHSF3KMDS","created_at":"2026-07-05T10:58:41.658369+00:00"},{"alias_kind":"pith_short_8","alias_value":"KOTFVC4K","created_at":"2026-07-05T10:58:41.658369+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07539","citing_title":"Prompt Governance? On Governing Technologies Governed by Natural Language","ref_index":21,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06708","citing_title":"Signal-Driven Observation for Long-Horizon Web Agents","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2512.00520","citing_title":"Toward a Safe Internet of Agents","ref_index":2,"is_internal_anchor":false},{"citing_arxiv_id":"2605.02542","citing_title":"IteRate: Autonomous AI Synthesis of In-Kernel eBPF Wi-Fi Rate Control Algorithms","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB","json":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB.json","graph_json":"https://pith.science/api/pith-number/KOTFVC4KHSF3KMDSLNGP4HO5AB/graph.json","events_json":"https://pith.science/api/pith-number/KOTFVC4KHSF3KMDSLNGP4HO5AB/events.json","paper":"https://pith.science/paper/KOTFVC4K"},"agent_actions":{"view_html":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB","download_json":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB.json","view_paper":"https://pith.science/paper/KOTFVC4K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.02709&json=true","fetch_graph":"https://pith.science/api/pith-number/KOTFVC4KHSF3KMDSLNGP4HO5AB/graph.json","fetch_events":"https://pith.science/api/pith-number/KOTFVC4KHSF3KMDSLNGP4HO5AB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB/action/storage_attestation","attest_author":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB/action/author_attestation","sign_citation":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB/action/citation_signature","submit_replication":"https://pith.science/pith/KOTFVC4KHSF3KMDSLNGP4HO5AB/action/replication_record"}},"created_at":"2026-07-05T10:58:41.658369+00:00","updated_at":"2026-07-05T10:58:41.658369+00:00"}