{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:QSJIIV5BWMQUK7L26S6QUBEOJM","short_pith_number":"pith:QSJIIV5B","schema_version":"1.0","canonical_sha256":"84928457a1b321457d7af4bd0a048e4b04d2fdd039f932b21ec4aec5b013ba0b","source":{"kind":"arxiv","id":"2607.09665","version":1},"attestation_state":"computed","paper":{"title":"Format Sensitivity Index: Token-Controlled Prompt Wrapper Robustness and Schema Compliance in LLM Benchmarking","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Deep Pankajbhai Mehta","submitted_at":"2026-05-02T22:51:41Z","abstract_excerpt":"Prompt wrappers often differ only in formatting, yet they can change model scores enough to flip leaderboard conclusions. We study this variance under a token-controlled protocol and introduce two complementary metrics: the Format Sensitivity Index (FSI), the accuracy range induced by wrapper choice, and the Parseability Sensitivity Index (PSI), the corresponding range in answer parseability. Across 140,000 OpenRouter generations spanning 7 QA tasks, 5 wrapper families, and 4 instruct models from 7B to 72B parameters, we find that mean FSI varies by over 30x across models and is largely explai"},"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":"2607.09665","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/publicdomain/zero/1.0/","primary_cat":"cs.AI","submitted_at":"2026-05-02T22:51:41Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"634af8ed72d289d544d6031a23e9511e5c95687f599f90afcf427fc1200a871d","abstract_canon_sha256":"5639c76a9163935f3d9378596c3b0dd18670f38a6315d80191bd324451c08dab"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-14T00:18:29.501547Z","signature_b64":"k26it/cAXwKpmXjaLeXtVphOFt0rXCd78G0+oPRju/H37npfs/6Ok0F9BV1rNJytr3N36XasDRtcOC66uYRlBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"84928457a1b321457d7af4bd0a048e4b04d2fdd039f932b21ec4aec5b013ba0b","last_reissued_at":"2026-07-14T00:18:29.500756Z","signature_status":"signed_v1","first_computed_at":"2026-07-14T00:18:29.500756Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Format Sensitivity Index: Token-Controlled Prompt Wrapper Robustness and Schema Compliance in LLM Benchmarking","license":"http://creativecommons.org/publicdomain/zero/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.AI","authors_text":"Deep Pankajbhai Mehta","submitted_at":"2026-05-02T22:51:41Z","abstract_excerpt":"Prompt wrappers often differ only in formatting, yet they can change model scores enough to flip leaderboard conclusions. We study this variance under a token-controlled protocol and introduce two complementary metrics: the Format Sensitivity Index (FSI), the accuracy range induced by wrapper choice, and the Parseability Sensitivity Index (PSI), the corresponding range in answer parseability. Across 140,000 OpenRouter generations spanning 7 QA tasks, 5 wrapper families, and 4 instruct models from 7B to 72B parameters, we find that mean FSI varies by over 30x across models and is largely explai"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.09665","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/2607.09665/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":"2607.09665","created_at":"2026-07-14T00:18:29.501158+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.09665v1","created_at":"2026-07-14T00:18:29.501158+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.09665","created_at":"2026-07-14T00:18:29.501158+00:00"},{"alias_kind":"pith_short_12","alias_value":"QSJIIV5BWMQU","created_at":"2026-07-14T00:18:29.501158+00:00"},{"alias_kind":"pith_short_16","alias_value":"QSJIIV5BWMQUK7L2","created_at":"2026-07-14T00:18:29.501158+00:00"},{"alias_kind":"pith_short_8","alias_value":"QSJIIV5B","created_at":"2026-07-14T00:18:29.501158+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/QSJIIV5BWMQUK7L26S6QUBEOJM","json":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM.json","graph_json":"https://pith.science/api/pith-number/QSJIIV5BWMQUK7L26S6QUBEOJM/graph.json","events_json":"https://pith.science/api/pith-number/QSJIIV5BWMQUK7L26S6QUBEOJM/events.json","paper":"https://pith.science/paper/QSJIIV5B"},"agent_actions":{"view_html":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM","download_json":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM.json","view_paper":"https://pith.science/paper/QSJIIV5B","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.09665&json=true","fetch_graph":"https://pith.science/api/pith-number/QSJIIV5BWMQUK7L26S6QUBEOJM/graph.json","fetch_events":"https://pith.science/api/pith-number/QSJIIV5BWMQUK7L26S6QUBEOJM/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM/action/storage_attestation","attest_author":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM/action/author_attestation","sign_citation":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM/action/citation_signature","submit_replication":"https://pith.science/pith/QSJIIV5BWMQUK7L26S6QUBEOJM/action/replication_record"}},"created_at":"2026-07-14T00:18:29.501158+00:00","updated_at":"2026-07-14T00:18:29.501158+00:00"}