{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:SX6F7LLLSTWW3NABL2XDWRYM6U","short_pith_number":"pith:SX6F7LLL","schema_version":"1.0","canonical_sha256":"95fc5fad6b94ed6db4015eae3b470cf50167147d39fb354c1250a3925fd04464","source":{"kind":"arxiv","id":"2606.32008","version":1},"attestation_state":"computed","paper":{"title":"Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aobo Yang, Ayush Warikoo, Chia-Tse Shao, Nehal Bandi, Philippe Chlenski, Vivek Miglani, Yingxiao Ye, Zachariah Carmichael","submitted_at":"2026-06-30T17:43:44Z","abstract_excerpt":"Mechanistic interpretability (MI) requires full access to model internals, yet the APIs for most widely deployed language models at best expose log-probabilities over output tokens. This creates a surrogate problem: when do measurements made on open models allow us to make claims about a closed model? We evaluate surrogate fidelity at the prediction, attribution, and representation levels. For binary classification tasks, log-odds provide an API-compatible scalar readout of the model's representation space, and leave-one-out attributions provide insight into model behavior. Across eleven model"},"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":"2606.32008","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.LG","submitted_at":"2026-06-30T17:43:44Z","cross_cats_sorted":[],"title_canon_sha256":"a4f8f93df4c32b48ff9c882f5d1926604f25814ee4e1bf22f5196d7e33aa6d31","abstract_canon_sha256":"a71bef9e850bc5ecd87a1be96d021484947e254abeeb8c50fdaecb07a6634057"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-01T01:18:28.939346Z","signature_b64":"7vwR1j0+Nxyd4tcvkWhLs9F8nT5kzAouIO32czFSCXCjHbRsXDRHIzTWWoOfB6CGaXJ+goUBMHqSKDzBCbn/Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"95fc5fad6b94ed6db4015eae3b470cf50167147d39fb354c1250a3925fd04464","last_reissued_at":"2026-07-01T01:18:28.938899Z","signature_status":"signed_v1","first_computed_at":"2026-07-01T01:18:28.938899Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Surrogate Fidelity: When Can Open LLMs Explain Closed Ones?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Aobo Yang, Ayush Warikoo, Chia-Tse Shao, Nehal Bandi, Philippe Chlenski, Vivek Miglani, Yingxiao Ye, Zachariah Carmichael","submitted_at":"2026-06-30T17:43:44Z","abstract_excerpt":"Mechanistic interpretability (MI) requires full access to model internals, yet the APIs for most widely deployed language models at best expose log-probabilities over output tokens. This creates a surrogate problem: when do measurements made on open models allow us to make claims about a closed model? We evaluate surrogate fidelity at the prediction, attribution, and representation levels. For binary classification tasks, log-odds provide an API-compatible scalar readout of the model's representation space, and leave-one-out attributions provide insight into model behavior. Across eleven model"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.32008","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/2606.32008/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":"2606.32008","created_at":"2026-07-01T01:18:28.938956+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.32008v1","created_at":"2026-07-01T01:18:28.938956+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.32008","created_at":"2026-07-01T01:18:28.938956+00:00"},{"alias_kind":"pith_short_12","alias_value":"SX6F7LLLSTWW","created_at":"2026-07-01T01:18:28.938956+00:00"},{"alias_kind":"pith_short_16","alias_value":"SX6F7LLLSTWW3NAB","created_at":"2026-07-01T01:18:28.938956+00:00"},{"alias_kind":"pith_short_8","alias_value":"SX6F7LLL","created_at":"2026-07-01T01:18:28.938956+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/SX6F7LLLSTWW3NABL2XDWRYM6U","json":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U.json","graph_json":"https://pith.science/api/pith-number/SX6F7LLLSTWW3NABL2XDWRYM6U/graph.json","events_json":"https://pith.science/api/pith-number/SX6F7LLLSTWW3NABL2XDWRYM6U/events.json","paper":"https://pith.science/paper/SX6F7LLL"},"agent_actions":{"view_html":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U","download_json":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U.json","view_paper":"https://pith.science/paper/SX6F7LLL","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.32008&json=true","fetch_graph":"https://pith.science/api/pith-number/SX6F7LLLSTWW3NABL2XDWRYM6U/graph.json","fetch_events":"https://pith.science/api/pith-number/SX6F7LLLSTWW3NABL2XDWRYM6U/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U/action/storage_attestation","attest_author":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U/action/author_attestation","sign_citation":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U/action/citation_signature","submit_replication":"https://pith.science/pith/SX6F7LLLSTWW3NABL2XDWRYM6U/action/replication_record"}},"created_at":"2026-07-01T01:18:28.938956+00:00","updated_at":"2026-07-01T01:18:28.938956+00:00"}