{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:URQGX36RMICXEZ7RTMITJVZ2OV","short_pith_number":"pith:URQGX36R","schema_version":"1.0","canonical_sha256":"a4606befd162057267f19b1134d73a754fcc3320ebf5e0bd5a2bde5920ecec21","source":{"kind":"arxiv","id":"2508.19980","version":1},"attestation_state":"computed","paper":{"title":"Evaluating Language Model Reasoning about Confidential Information","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Robey, Andy Zou, Dylan Sam, J. Zico Kolter, Matt Fredrikson","submitted_at":"2025-08-27T15:39:46Z","abstract_excerpt":"As language models are increasingly deployed as autonomous agents in high-stakes settings, ensuring that they reliably follow user-defined rules has become a critical safety concern. To this end, we study whether language models exhibit contextual robustness, or the capability to adhere to context-dependent safety specifications. For this analysis, we develop a benchmark (PasswordEval) that measures whether language models can correctly determine when a user request is authorized (i.e., with a correct password). We find that current open- and closed-source models struggle with this seemingly s"},"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":"2508.19980","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-08-27T15:39:46Z","cross_cats_sorted":[],"title_canon_sha256":"48cdc9f8a2db77d92ba3e746711c7231cdb0304bcc2a2201c8b3eb6f9e1bc312","abstract_canon_sha256":"e25c8c3d37918987ec835d8cad138fa9629a1e66f593cdfe2ae583f52a9f8aff"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T12:00:21.176215Z","signature_b64":"TvpgmLUl0JNVKw5pU53x23dMX68CiVzQ6BR90/FYnF31N/0ni2eKPOzJFSMl+fj/p4iKj5ZNDLLsGbxlpvTfDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"a4606befd162057267f19b1134d73a754fcc3320ebf5e0bd5a2bde5920ecec21","last_reissued_at":"2026-07-05T12:00:21.175827Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T12:00:21.175827Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Evaluating Language Model Reasoning about Confidential Information","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alexander Robey, Andy Zou, Dylan Sam, J. Zico Kolter, Matt Fredrikson","submitted_at":"2025-08-27T15:39:46Z","abstract_excerpt":"As language models are increasingly deployed as autonomous agents in high-stakes settings, ensuring that they reliably follow user-defined rules has become a critical safety concern. To this end, we study whether language models exhibit contextual robustness, or the capability to adhere to context-dependent safety specifications. For this analysis, we develop a benchmark (PasswordEval) that measures whether language models can correctly determine when a user request is authorized (i.e., with a correct password). We find that current open- and closed-source models struggle with this seemingly s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2508.19980","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/2508.19980/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":"2508.19980","created_at":"2026-07-05T12:00:21.175888+00:00"},{"alias_kind":"arxiv_version","alias_value":"2508.19980v1","created_at":"2026-07-05T12:00:21.175888+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2508.19980","created_at":"2026-07-05T12:00:21.175888+00:00"},{"alias_kind":"pith_short_12","alias_value":"URQGX36RMICX","created_at":"2026-07-05T12:00:21.175888+00:00"},{"alias_kind":"pith_short_16","alias_value":"URQGX36RMICXEZ7R","created_at":"2026-07-05T12:00:21.175888+00:00"},{"alias_kind":"pith_short_8","alias_value":"URQGX36R","created_at":"2026-07-05T12:00:21.175888+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/URQGX36RMICXEZ7RTMITJVZ2OV","json":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV.json","graph_json":"https://pith.science/api/pith-number/URQGX36RMICXEZ7RTMITJVZ2OV/graph.json","events_json":"https://pith.science/api/pith-number/URQGX36RMICXEZ7RTMITJVZ2OV/events.json","paper":"https://pith.science/paper/URQGX36R"},"agent_actions":{"view_html":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV","download_json":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV.json","view_paper":"https://pith.science/paper/URQGX36R","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2508.19980&json=true","fetch_graph":"https://pith.science/api/pith-number/URQGX36RMICXEZ7RTMITJVZ2OV/graph.json","fetch_events":"https://pith.science/api/pith-number/URQGX36RMICXEZ7RTMITJVZ2OV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV/action/storage_attestation","attest_author":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV/action/author_attestation","sign_citation":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV/action/citation_signature","submit_replication":"https://pith.science/pith/URQGX36RMICXEZ7RTMITJVZ2OV/action/replication_record"}},"created_at":"2026-07-05T12:00:21.175888+00:00","updated_at":"2026-07-05T12:00:21.175888+00:00"}