{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YFC6SQ5HHGUZILZ24H7UW3XCBG","short_pith_number":"pith:YFC6SQ5H","schema_version":"1.0","canonical_sha256":"c145e943a739a9942f3ae1ff4b6ee2098f924e05c4c1736adc75d13453b7af6d","source":{"kind":"arxiv","id":"2506.06971","version":2},"attestation_state":"computed","paper":{"title":"Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Arin Garg, Jaechul Roh, Shivani Anilkumar, Varun Gandhi","submitted_at":"2025-06-08T02:43:46Z","abstract_excerpt":"Large Language Models (LLMs) have achieved remarkable success in tasks requiring complex reasoning, such as code generation, mathematical problem solving, and algorithmic synthesis -- especially when aided by reasoning tokens and Chain-of-Thought prompting. Yet, a core question remains: do these models truly reason, or do they merely exploit shallow statistical patterns? In this paper, we introduce Chain-of-Code Collapse, where we systematically investigate the robustness of reasoning LLMs by introducing a suite of semantically faithful yet adversarially structured prompt perturbations. Our ev"},"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":"2506.06971","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-06-08T02:43:46Z","cross_cats_sorted":["cs.CR"],"title_canon_sha256":"314298158eafb7dea60d315f1c262480b44e6cbb6ebcc5cbe949f95779824f5a","abstract_canon_sha256":"dbe978e6c3e5bda02eb43d9f7da2dc6f08bd4a28043d39bb1c0a1c33d8c36f42"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:20:13.762263Z","signature_b64":"xQt5BAJ/gvBFb0LjPPPkECwOqH2SUyzVP6C6yV2keI+Xo3Y+BmyJ6KIJBBGl7mhFuakQBJ7ZEliTp3lp05n8AQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c145e943a739a9942f3ae1ff4b6ee2098f924e05c4c1736adc75d13453b7af6d","last_reissued_at":"2026-07-05T11:20:13.761728Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:20:13.761728Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CR"],"primary_cat":"cs.CL","authors_text":"Arin Garg, Jaechul Roh, Shivani Anilkumar, Varun Gandhi","submitted_at":"2025-06-08T02:43:46Z","abstract_excerpt":"Large Language Models (LLMs) have achieved remarkable success in tasks requiring complex reasoning, such as code generation, mathematical problem solving, and algorithmic synthesis -- especially when aided by reasoning tokens and Chain-of-Thought prompting. Yet, a core question remains: do these models truly reason, or do they merely exploit shallow statistical patterns? In this paper, we introduce Chain-of-Code Collapse, where we systematically investigate the robustness of reasoning LLMs by introducing a suite of semantically faithful yet adversarially structured prompt perturbations. Our ev"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.06971","kind":"arxiv","version":2},"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/2506.06971/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":"2506.06971","created_at":"2026-07-05T11:20:13.761791+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.06971v2","created_at":"2026-07-05T11:20:13.761791+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.06971","created_at":"2026-07-05T11:20:13.761791+00:00"},{"alias_kind":"pith_short_12","alias_value":"YFC6SQ5HHGUZ","created_at":"2026-07-05T11:20:13.761791+00:00"},{"alias_kind":"pith_short_16","alias_value":"YFC6SQ5HHGUZILZ2","created_at":"2026-07-05T11:20:13.761791+00:00"},{"alias_kind":"pith_short_8","alias_value":"YFC6SQ5H","created_at":"2026-07-05T11:20:13.761791+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":6,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.03606","citing_title":"Testing LLM Arithmetic Reasoning Generalization with Automatic Numeric-Remapping Attacks","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2512.05929","citing_title":"LLM Harms: A Taxonomy and Discussion","ref_index":104,"is_internal_anchor":false},{"citing_arxiv_id":"2603.03332","citing_title":"Fragile Thoughts: How Large Language Models Handle Chain-of-Thought Perturbations","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02776","citing_title":"Evaluating the Environmental Impact of using SLMs and Prompt Engineering for Code Generation","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2604.25247","citing_title":"R-CoT: A Reasoning-Layer Watermark via Redundant Chain-of-Thought in Large Language Models","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2604.24712","citing_title":"When Prompt Under-Specification Improves Code Correctness: An Exploratory Study of Prompt Wording and Structure Effects on LLM-Based Code Generation","ref_index":5,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG","json":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG.json","graph_json":"https://pith.science/api/pith-number/YFC6SQ5HHGUZILZ24H7UW3XCBG/graph.json","events_json":"https://pith.science/api/pith-number/YFC6SQ5HHGUZILZ24H7UW3XCBG/events.json","paper":"https://pith.science/paper/YFC6SQ5H"},"agent_actions":{"view_html":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG","download_json":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG.json","view_paper":"https://pith.science/paper/YFC6SQ5H","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.06971&json=true","fetch_graph":"https://pith.science/api/pith-number/YFC6SQ5HHGUZILZ24H7UW3XCBG/graph.json","fetch_events":"https://pith.science/api/pith-number/YFC6SQ5HHGUZILZ24H7UW3XCBG/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG/action/storage_attestation","attest_author":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG/action/author_attestation","sign_citation":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG/action/citation_signature","submit_replication":"https://pith.science/pith/YFC6SQ5HHGUZILZ24H7UW3XCBG/action/replication_record"}},"created_at":"2026-07-05T11:20:13.761791+00:00","updated_at":"2026-07-05T11:20:13.761791+00:00"}