{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:H3INTHIPUHEGNDNSG7G66J2I3F","short_pith_number":"pith:H3INTHIP","schema_version":"1.0","canonical_sha256":"3ed0d99d0fa1c8668db237cdef2748d944b197f6af3b8ca2689dfa086e1b1ce9","source":{"kind":"arxiv","id":"2603.16475","version":2},"attestation_state":"computed","paper":{"title":"Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alexander Panchenko, Elena Tutubalina, Gleb Ershov, Karim Vafin, Mikhail Chaichuk, Mikhail Seleznyov, Oleg Somov","submitted_at":"2026-03-17T13:01:44Z","abstract_excerpt":"In schema-guided reasoning (SGR) pipelines, LLMs produce explicit intermediate structures -- rubrics, checklists, or verification queries -- before committing to a final decision. SGR is increasingly adopted because it promises controllability: practitioners expect to inspect, edit, and override these structures to steer the outcome. But does the promise hold? We introduce a causal evaluation protocol to measure it: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across 12 models and 4 benchmark"},"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":"2603.16475","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2026-03-17T13:01:44Z","cross_cats_sorted":[],"title_canon_sha256":"9054e6690d0e3d1aa7320ce8ecc399e2456b12367b1700861f07280851f5e61e","abstract_canon_sha256":"db248779c638cadcbed5678dd6945e2dbb33fe68e088f3bf8f601ddf082c4acd"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-05T01:14:36.772448Z","signature_b64":"upuF0k0MUxMxAJhgmby1Jy3XUX5kui8EGJMa6lplHRbOKW1UilQZ5j9eaZHJLxEyHw603tEVE+Sv0KTia2DBBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"3ed0d99d0fa1c8668db237cdef2748d944b197f6af3b8ca2689dfa086e1b1ce9","last_reissued_at":"2026-06-05T01:14:36.771679Z","signature_status":"signed_v1","first_computed_at":"2026-06-05T01:14:36.771679Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Breaking the Chain: A Causal Analysis of LLM Faithfulness to Intermediate Structures","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Alexander Panchenko, Elena Tutubalina, Gleb Ershov, Karim Vafin, Mikhail Chaichuk, Mikhail Seleznyov, Oleg Somov","submitted_at":"2026-03-17T13:01:44Z","abstract_excerpt":"In schema-guided reasoning (SGR) pipelines, LLMs produce explicit intermediate structures -- rubrics, checklists, or verification queries -- before committing to a final decision. SGR is increasingly adopted because it promises controllability: practitioners expect to inspect, edit, and override these structures to steer the outcome. But does the promise hold? We introduce a causal evaluation protocol to measure it: by selecting tasks where a deterministic function maps intermediate structures to decisions, every controlled edit implies a unique correct output. Across 12 models and 4 benchmark"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2603.16475","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/2603.16475/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":"2603.16475","created_at":"2026-06-05T01:14:36.771797+00:00"},{"alias_kind":"arxiv_version","alias_value":"2603.16475v2","created_at":"2026-06-05T01:14:36.771797+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2603.16475","created_at":"2026-06-05T01:14:36.771797+00:00"},{"alias_kind":"pith_short_12","alias_value":"H3INTHIPUHEG","created_at":"2026-06-05T01:14:36.771797+00:00"},{"alias_kind":"pith_short_16","alias_value":"H3INTHIPUHEGNDNS","created_at":"2026-06-05T01:14:36.771797+00:00"},{"alias_kind":"pith_short_8","alias_value":"H3INTHIP","created_at":"2026-06-05T01:14:36.771797+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/H3INTHIPUHEGNDNSG7G66J2I3F","json":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F.json","graph_json":"https://pith.science/api/pith-number/H3INTHIPUHEGNDNSG7G66J2I3F/graph.json","events_json":"https://pith.science/api/pith-number/H3INTHIPUHEGNDNSG7G66J2I3F/events.json","paper":"https://pith.science/paper/H3INTHIP"},"agent_actions":{"view_html":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F","download_json":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F.json","view_paper":"https://pith.science/paper/H3INTHIP","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2603.16475&json=true","fetch_graph":"https://pith.science/api/pith-number/H3INTHIPUHEGNDNSG7G66J2I3F/graph.json","fetch_events":"https://pith.science/api/pith-number/H3INTHIPUHEGNDNSG7G66J2I3F/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F/action/timestamp_anchor","attest_storage":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F/action/storage_attestation","attest_author":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F/action/author_attestation","sign_citation":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F/action/citation_signature","submit_replication":"https://pith.science/pith/H3INTHIPUHEGNDNSG7G66J2I3F/action/replication_record"}},"created_at":"2026-06-05T01:14:36.771797+00:00","updated_at":"2026-06-05T01:14:36.771797+00:00"}