{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TELMW3TK4AFFTNGOLV6A2G4ANO","short_pith_number":"pith:TELMW3TK","schema_version":"1.0","canonical_sha256":"9916cb6e6ae00a59b4ce5d7c0d1b806b8df9ac73643780d24e7b48f75a582412","source":{"kind":"arxiv","id":"2606.01148","version":1},"attestation_state":"computed","paper":{"title":"Not All Explanations Simulate Equally: Comparing Verbalized Feature Attributions and Self-Generated Rationales","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Roth, Pingjun Hong","submitted_at":"2026-05-31T10:35:35Z","abstract_excerpt":"Natural-language explanations are often treated as a unified interface for understanding model behavior, but different explanation sources may support simulation in different ways. This paper compares two families of explanations for question answering models: verbalized feature attributions and self-generated rationales. We evaluate them under a shared counterfactual simulation setting, using an LLM judge as predictor and measuring whether it can better predict a model's answers to follow-up questions when given its explanation. Across multiple instruction-tuned models, we analyze how explana"},"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":"2606.01148","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-31T10:35:35Z","cross_cats_sorted":[],"title_canon_sha256":"c93e948041681a33d8d5c78a9fdf5160e009b455f2d832daffef0b3e8e2e9af6","abstract_canon_sha256":"c62e91ae300f17f078a468ff15733410333c02b54e1cf61c3e316ab8f0ac1fa3"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T02:04:25.083831Z","signature_b64":"UVnyNi103rRUVTr1aUqZUBAqyCqUvrPUO1NJxutBXd1llR2xWSokDBcQwU9wRguqAw5pZGmcreC99JBmMUeLDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9916cb6e6ae00a59b4ce5d7c0d1b806b8df9ac73643780d24e7b48f75a582412","last_reissued_at":"2026-06-02T02:04:25.083403Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T02:04:25.083403Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Not All Explanations Simulate Equally: Comparing Verbalized Feature Attributions and Self-Generated Rationales","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Benjamin Roth, Pingjun Hong","submitted_at":"2026-05-31T10:35:35Z","abstract_excerpt":"Natural-language explanations are often treated as a unified interface for understanding model behavior, but different explanation sources may support simulation in different ways. This paper compares two families of explanations for question answering models: verbalized feature attributions and self-generated rationales. We evaluate them under a shared counterfactual simulation setting, using an LLM judge as predictor and measuring whether it can better predict a model's answers to follow-up questions when given its explanation. Across multiple instruction-tuned models, we analyze how explana"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.01148","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/2606.01148/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":"2606.01148","created_at":"2026-06-02T02:04:25.083464+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.01148v1","created_at":"2026-06-02T02:04:25.083464+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.01148","created_at":"2026-06-02T02:04:25.083464+00:00"},{"alias_kind":"pith_short_12","alias_value":"TELMW3TK4AFF","created_at":"2026-06-02T02:04:25.083464+00:00"},{"alias_kind":"pith_short_16","alias_value":"TELMW3TK4AFFTNGO","created_at":"2026-06-02T02:04:25.083464+00:00"},{"alias_kind":"pith_short_8","alias_value":"TELMW3TK","created_at":"2026-06-02T02:04:25.083464+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/TELMW3TK4AFFTNGOLV6A2G4ANO","json":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO.json","graph_json":"https://pith.science/api/pith-number/TELMW3TK4AFFTNGOLV6A2G4ANO/graph.json","events_json":"https://pith.science/api/pith-number/TELMW3TK4AFFTNGOLV6A2G4ANO/events.json","paper":"https://pith.science/paper/TELMW3TK"},"agent_actions":{"view_html":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO","download_json":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO.json","view_paper":"https://pith.science/paper/TELMW3TK","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.01148&json=true","fetch_graph":"https://pith.science/api/pith-number/TELMW3TK4AFFTNGOLV6A2G4ANO/graph.json","fetch_events":"https://pith.science/api/pith-number/TELMW3TK4AFFTNGOLV6A2G4ANO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO/action/storage_attestation","attest_author":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO/action/author_attestation","sign_citation":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO/action/citation_signature","submit_replication":"https://pith.science/pith/TELMW3TK4AFFTNGOLV6A2G4ANO/action/replication_record"}},"created_at":"2026-06-02T02:04:25.083464+00:00","updated_at":"2026-06-02T02:04:25.083464+00:00"}