{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:RTOKY3YBFZ57UOZFRBUVTSBGVF","short_pith_number":"pith:RTOKY3YB","schema_version":"1.0","canonical_sha256":"8cdcac6f012e7bfa3b25886959c826a94cac33fd095128034a13e40c2cd60da9","source":{"kind":"arxiv","id":"2608.02639","version":1},"attestation_state":"computed","paper":{"title":"Instruction Stacking Collapse: A Benchmark and the Capability-Dependent Value of Prompt Compilation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Atul Anand, Sourav Chattaraj","submitted_at":"2026-07-31T10:58:29Z","abstract_excerpt":"Production prompts rarely carry a single instruction. One system message may require valid JSON, a word limit, three citations, and a fixed tone at the same time. We study how instruction-following degrades as such constraints accumulate. We introduce a benchmark that stacks 24 verifier-checked instructions, one to twenty at a time, and evaluate three production-tier LLMs (Claude Sonnet 4.6, GPT-5-mini, Gemini 2.5 Flash). Instruction-following degrades non-linearly: the follow rate falls from ~96% to as low as 20%, driven by a structured and reproducible set of pairwise conflicts. A single \"ou"},"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":"2608.02639","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.SE","submitted_at":"2026-07-31T10:58:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"ad2d046a97d4073d4cdcfc5cb5749e6aa37820a0ac4ccedef91818732a4b6364","abstract_canon_sha256":"eb1ef4500eb6ace056bfb1e44ce10d1fe5a8482088d1da9dfb54cd863045c870"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-08-05T00:38:05.690297Z","signature_b64":"InO9MzowXT2qoVWYXiGGga5g4GoQ+f1JLk9jLoJ6HL5TqeET7zjQmlAaKXyRilfcuK7ROMvOsR5UqH7RmXUHBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"8cdcac6f012e7bfa3b25886959c826a94cac33fd095128034a13e40c2cd60da9","last_reissued_at":"2026-08-05T00:38:05.687782Z","signature_status":"signed_v1","first_computed_at":"2026-08-05T00:38:05.687782Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Instruction Stacking Collapse: A Benchmark and the Capability-Dependent Value of Prompt Compilation","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.SE","authors_text":"Atul Anand, Sourav Chattaraj","submitted_at":"2026-07-31T10:58:29Z","abstract_excerpt":"Production prompts rarely carry a single instruction. One system message may require valid JSON, a word limit, three citations, and a fixed tone at the same time. We study how instruction-following degrades as such constraints accumulate. We introduce a benchmark that stacks 24 verifier-checked instructions, one to twenty at a time, and evaluate three production-tier LLMs (Claude Sonnet 4.6, GPT-5-mini, Gemini 2.5 Flash). Instruction-following degrades non-linearly: the follow rate falls from ~96% to as low as 20%, driven by a structured and reproducible set of pairwise conflicts. A single \"ou"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2608.02639","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/2608.02639/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":"2608.02639","created_at":"2026-08-05T00:38:05.688704+00:00"},{"alias_kind":"arxiv_version","alias_value":"2608.02639v1","created_at":"2026-08-05T00:38:05.688704+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2608.02639","created_at":"2026-08-05T00:38:05.688704+00:00"},{"alias_kind":"pith_short_12","alias_value":"RTOKY3YBFZ57","created_at":"2026-08-05T00:38:05.688704+00:00"},{"alias_kind":"pith_short_16","alias_value":"RTOKY3YBFZ57UOZF","created_at":"2026-08-05T00:38:05.688704+00:00"},{"alias_kind":"pith_short_8","alias_value":"RTOKY3YB","created_at":"2026-08-05T00:38:05.688704+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/RTOKY3YBFZ57UOZFRBUVTSBGVF","json":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF.json","graph_json":"https://pith.science/api/pith-number/RTOKY3YBFZ57UOZFRBUVTSBGVF/graph.json","events_json":"https://pith.science/api/pith-number/RTOKY3YBFZ57UOZFRBUVTSBGVF/events.json","paper":"https://pith.science/paper/RTOKY3YB"},"agent_actions":{"view_html":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF","download_json":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF.json","view_paper":"https://pith.science/paper/RTOKY3YB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2608.02639&json=true","fetch_graph":"https://pith.science/api/pith-number/RTOKY3YBFZ57UOZFRBUVTSBGVF/graph.json","fetch_events":"https://pith.science/api/pith-number/RTOKY3YBFZ57UOZFRBUVTSBGVF/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF/action/timestamp_anchor","attest_storage":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF/action/storage_attestation","attest_author":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF/action/author_attestation","sign_citation":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF/action/citation_signature","submit_replication":"https://pith.science/pith/RTOKY3YBFZ57UOZFRBUVTSBGVF/action/replication_record"}},"created_at":"2026-08-05T00:38:05.688704+00:00","updated_at":"2026-08-05T00:38:05.688704+00:00"}