{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:E2ZP7Z4KQEBBICBDRCCILDTFXB","short_pith_number":"pith:E2ZP7Z4K","schema_version":"1.0","canonical_sha256":"26b2ffe78a81021408238884858e65b87b3986af1aa1551f5e3c76512ac4e2f2","source":{"kind":"arxiv","id":"2506.20471","version":1},"attestation_state":"computed","paper":{"title":"Probing AI Safety with Source Code","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ameet Deshpande, Ashwin Kalyan, Karthik Narasimhan, Shreyas Chaudhari, Tanmay Rajpurohit, Ujwal Narayan, Vishvak Murahari","submitted_at":"2025-06-25T14:19:57Z","abstract_excerpt":"Large language models (LLMs) have become ubiquitous, interfacing with humans in numerous safety-critical applications. This necessitates improving capabilities, but importantly coupled with greater safety measures to align these models with human values and preferences. In this work, we demonstrate that contemporary models fall concerningly short of the goal of AI safety, leading to an unsafe and harmful experience for users. We introduce a prompting strategy called Code of Thought (CoDoT) to evaluate the safety of LLMs. CoDoT converts natural language inputs to simple code that represents the"},"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":"2506.20471","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-25T14:19:57Z","cross_cats_sorted":[],"title_canon_sha256":"b2a525633973140fc4bf77dbfbda125fa90d7c8100421f70b1fef81c6ab4691a","abstract_canon_sha256":"4b9e6f406c4749dde2c39677172fcc75650f7f76081306371b06e13186ed874d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:27:06.308792Z","signature_b64":"M7K3HFOt2E52i22s19+7jeIB8kLuxWaNxd4GahC9YGoZBI6vqYE3fVoVVWTBa/umJuXu1UnaRG869xcDgawtBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"26b2ffe78a81021408238884858e65b87b3986af1aa1551f5e3c76512ac4e2f2","last_reissued_at":"2026-07-05T11:27:06.308329Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:27:06.308329Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Probing AI Safety with Source Code","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Ameet Deshpande, Ashwin Kalyan, Karthik Narasimhan, Shreyas Chaudhari, Tanmay Rajpurohit, Ujwal Narayan, Vishvak Murahari","submitted_at":"2025-06-25T14:19:57Z","abstract_excerpt":"Large language models (LLMs) have become ubiquitous, interfacing with humans in numerous safety-critical applications. This necessitates improving capabilities, but importantly coupled with greater safety measures to align these models with human values and preferences. In this work, we demonstrate that contemporary models fall concerningly short of the goal of AI safety, leading to an unsafe and harmful experience for users. We introduce a prompting strategy called Code of Thought (CoDoT) to evaluate the safety of LLMs. CoDoT converts natural language inputs to simple code that represents the"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.20471","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/2506.20471/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":"2506.20471","created_at":"2026-07-05T11:27:06.308390+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.20471v1","created_at":"2026-07-05T11:27:06.308390+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.20471","created_at":"2026-07-05T11:27:06.308390+00:00"},{"alias_kind":"pith_short_12","alias_value":"E2ZP7Z4KQEBB","created_at":"2026-07-05T11:27:06.308390+00:00"},{"alias_kind":"pith_short_16","alias_value":"E2ZP7Z4KQEBBICBD","created_at":"2026-07-05T11:27:06.308390+00:00"},{"alias_kind":"pith_short_8","alias_value":"E2ZP7Z4K","created_at":"2026-07-05T11:27:06.308390+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/E2ZP7Z4KQEBBICBDRCCILDTFXB","json":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB.json","graph_json":"https://pith.science/api/pith-number/E2ZP7Z4KQEBBICBDRCCILDTFXB/graph.json","events_json":"https://pith.science/api/pith-number/E2ZP7Z4KQEBBICBDRCCILDTFXB/events.json","paper":"https://pith.science/paper/E2ZP7Z4K"},"agent_actions":{"view_html":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB","download_json":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB.json","view_paper":"https://pith.science/paper/E2ZP7Z4K","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.20471&json=true","fetch_graph":"https://pith.science/api/pith-number/E2ZP7Z4KQEBBICBDRCCILDTFXB/graph.json","fetch_events":"https://pith.science/api/pith-number/E2ZP7Z4KQEBBICBDRCCILDTFXB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB/action/storage_attestation","attest_author":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB/action/author_attestation","sign_citation":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB/action/citation_signature","submit_replication":"https://pith.science/pith/E2ZP7Z4KQEBBICBDRCCILDTFXB/action/replication_record"}},"created_at":"2026-07-05T11:27:06.308390+00:00","updated_at":"2026-07-05T11:27:06.308390+00:00"}