{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:BGC2WEQEFPJYTROMERN6GETJWR","short_pith_number":"pith:BGC2WEQE","schema_version":"1.0","canonical_sha256":"0985ab12042bd389c5cc245be31269b464bb314b0543b33767a4d37e1a783398","source":{"kind":"arxiv","id":"2507.06445","version":1},"attestation_state":"computed","paper":{"title":"Can Interpretation Predict Behavior on Unseen Data?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"David Alvarez-Melis, Jenny Kaufmann, Martin Wattenberg, Naomi Saphra, Victoria R. Li","submitted_at":"2025-07-08T23:07:33Z","abstract_excerpt":"Interpretability research often aims to predict how a model will respond to targeted interventions on specific mechanisms. However, it rarely predicts how a model will respond to unseen input data. This paper explores the promises and challenges of interpretability as a tool for predicting out-of-distribution (OOD) model behavior. Specifically, we investigate the correspondence between attention patterns and OOD generalization in hundreds of Transformer models independently trained on a synthetic classification task. These models exhibit several distinct systematic generalization rules OOD, fo"},"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":"2507.06445","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-08T23:07:33Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"4bf4b6e128182faf4d96709bfb5bde8e14a9864c5e97c19a885279bf20a0372a","abstract_canon_sha256":"e6d8993343f5b7f413c16d44151ca6ed16d52af20bae5d2e7c7276e86e14abce"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:34:15.954668Z","signature_b64":"Br0x/7N/tzAa2hRxcS6pL4EXUYlSR6nQToRIp1gI5exkgckyXKPt+MiuY0U8fKVWb0j+j9zipwcZhxIKiHN4Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"0985ab12042bd389c5cc245be31269b464bb314b0543b33767a4d37e1a783398","last_reissued_at":"2026-07-05T11:34:15.954148Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:34:15.954148Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can Interpretation Predict Behavior on Unseen Data?","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"David Alvarez-Melis, Jenny Kaufmann, Martin Wattenberg, Naomi Saphra, Victoria R. Li","submitted_at":"2025-07-08T23:07:33Z","abstract_excerpt":"Interpretability research often aims to predict how a model will respond to targeted interventions on specific mechanisms. However, it rarely predicts how a model will respond to unseen input data. This paper explores the promises and challenges of interpretability as a tool for predicting out-of-distribution (OOD) model behavior. Specifically, we investigate the correspondence between attention patterns and OOD generalization in hundreds of Transformer models independently trained on a synthetic classification task. These models exhibit several distinct systematic generalization rules OOD, fo"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.06445","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/2507.06445/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":"2507.06445","created_at":"2026-07-05T11:34:15.954206+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.06445v1","created_at":"2026-07-05T11:34:15.954206+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.06445","created_at":"2026-07-05T11:34:15.954206+00:00"},{"alias_kind":"pith_short_12","alias_value":"BGC2WEQEFPJY","created_at":"2026-07-05T11:34:15.954206+00:00"},{"alias_kind":"pith_short_16","alias_value":"BGC2WEQEFPJYTROM","created_at":"2026-07-05T11:34:15.954206+00:00"},{"alias_kind":"pith_short_8","alias_value":"BGC2WEQE","created_at":"2026-07-05T11:34:15.954206+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"2606.19317","citing_title":"Explaining Attention with Program Synthesis","ref_index":20,"is_internal_anchor":true},{"citing_arxiv_id":"2606.19317","citing_title":"Explaining Attention with Program Synthesis","ref_index":20,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR","json":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR.json","graph_json":"https://pith.science/api/pith-number/BGC2WEQEFPJYTROMERN6GETJWR/graph.json","events_json":"https://pith.science/api/pith-number/BGC2WEQEFPJYTROMERN6GETJWR/events.json","paper":"https://pith.science/paper/BGC2WEQE"},"agent_actions":{"view_html":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR","download_json":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR.json","view_paper":"https://pith.science/paper/BGC2WEQE","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.06445&json=true","fetch_graph":"https://pith.science/api/pith-number/BGC2WEQEFPJYTROMERN6GETJWR/graph.json","fetch_events":"https://pith.science/api/pith-number/BGC2WEQEFPJYTROMERN6GETJWR/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR/action/storage_attestation","attest_author":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR/action/author_attestation","sign_citation":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR/action/citation_signature","submit_replication":"https://pith.science/pith/BGC2WEQEFPJYTROMERN6GETJWR/action/replication_record"}},"created_at":"2026-07-05T11:34:15.954206+00:00","updated_at":"2026-07-05T11:34:15.954206+00:00"}