{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:HCWU5YICHTAMSAQLYNFDWIGDR6","short_pith_number":"pith:HCWU5YIC","schema_version":"1.0","canonical_sha256":"38ad4ee1023cc0c9020bc34a3b20c38fa8c6296eb224bb69b890a9b733433e63","source":{"kind":"arxiv","id":"2607.01153","version":1},"attestation_state":"computed","paper":{"title":"Adversarial Pragmatics for AI Safety Evaluation: A Benchmark for Instruction Conflict, Embedded Commands, and Policy Ambiguity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Brett Reynolds","submitted_at":"2026-07-01T16:33:14Z","abstract_excerpt":"Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model has followed an instruction, refused appropriately, complied with a policy, resisted an embedded command, or misreported progress in an agentic task. Existing benchmarks often compress these distinctions into pass/fail labels, obscuring whether failures arise from capability limits, policy ambiguity, instruction conflict, scaffold failure, or unstable evaluator judgments.\n  This paper introduces adversarial pragmatics as a benchmark and annotation protocol for eva"},"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":"2607.01153","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-07-01T16:33:14Z","cross_cats_sorted":["cs.AI","cs.SE"],"title_canon_sha256":"d16502cd6ab74e3792c26a6579512f2cd05c74a432a29075ef61f2f24dc9a539","abstract_canon_sha256":"5ee8d8078c7c86c37482d57bd1787000f523ebbbb391ddd31af33832b9deac8f"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-02T01:18:30.632838Z","signature_b64":"3X14Oe0B1GI/0BdtWgF3ZgsPUKlC8S+gIrPXXZLy/HDvihOIxu92lR9Etm019bzAuLh6inDv9zlge9SAES0sCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"38ad4ee1023cc0c9020bc34a3b20c38fa8c6296eb224bb69b890a9b733433e63","last_reissued_at":"2026-07-02T01:18:30.632435Z","signature_status":"signed_v1","first_computed_at":"2026-07-02T01:18:30.632435Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Adversarial Pragmatics for AI Safety Evaluation: A Benchmark for Instruction Conflict, Embedded Commands, and Policy Ambiguity","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.SE"],"primary_cat":"cs.CL","authors_text":"Brett Reynolds","submitted_at":"2026-07-01T16:33:14Z","abstract_excerpt":"Safety evaluations for language models increasingly depend on judgments about ambiguous natural-language behaviour: whether a model has followed an instruction, refused appropriately, complied with a policy, resisted an embedded command, or misreported progress in an agentic task. Existing benchmarks often compress these distinctions into pass/fail labels, obscuring whether failures arise from capability limits, policy ambiguity, instruction conflict, scaffold failure, or unstable evaluator judgments.\n  This paper introduces adversarial pragmatics as a benchmark and annotation protocol for eva"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.01153","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/2607.01153/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":"2607.01153","created_at":"2026-07-02T01:18:30.632490+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.01153v1","created_at":"2026-07-02T01:18:30.632490+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.01153","created_at":"2026-07-02T01:18:30.632490+00:00"},{"alias_kind":"pith_short_12","alias_value":"HCWU5YICHTAM","created_at":"2026-07-02T01:18:30.632490+00:00"},{"alias_kind":"pith_short_16","alias_value":"HCWU5YICHTAMSAQL","created_at":"2026-07-02T01:18:30.632490+00:00"},{"alias_kind":"pith_short_8","alias_value":"HCWU5YIC","created_at":"2026-07-02T01:18:30.632490+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/HCWU5YICHTAMSAQLYNFDWIGDR6","json":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6.json","graph_json":"https://pith.science/api/pith-number/HCWU5YICHTAMSAQLYNFDWIGDR6/graph.json","events_json":"https://pith.science/api/pith-number/HCWU5YICHTAMSAQLYNFDWIGDR6/events.json","paper":"https://pith.science/paper/HCWU5YIC"},"agent_actions":{"view_html":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6","download_json":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6.json","view_paper":"https://pith.science/paper/HCWU5YIC","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.01153&json=true","fetch_graph":"https://pith.science/api/pith-number/HCWU5YICHTAMSAQLYNFDWIGDR6/graph.json","fetch_events":"https://pith.science/api/pith-number/HCWU5YICHTAMSAQLYNFDWIGDR6/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6/action/timestamp_anchor","attest_storage":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6/action/storage_attestation","attest_author":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6/action/author_attestation","sign_citation":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6/action/citation_signature","submit_replication":"https://pith.science/pith/HCWU5YICHTAMSAQLYNFDWIGDR6/action/replication_record"}},"created_at":"2026-07-02T01:18:30.632490+00:00","updated_at":"2026-07-02T01:18:30.632490+00:00"}