{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:PRSNTZXXXDZSW3JFKLBK5JKO7Y","short_pith_number":"pith:PRSNTZXX","schema_version":"1.0","canonical_sha256":"7c64d9e6f7b8f32b6d2552c2aea54efe37003274aabd29b951abc7e71c8ab583","source":{"kind":"arxiv","id":"2505.14424","version":1},"attestation_state":"computed","paper":{"title":"Explaining Neural Networks with Reasons","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hannes Leitgeb, Levin Hornischer","submitted_at":"2025-05-20T14:32:03Z","abstract_excerpt":"We propose a new interpretability method for neural networks, which is based on a novel mathematico-philosophical theory of reasons. Our method computes a vector for each neuron, called its reasons vector. We then can compute how strongly this reasons vector speaks for various propositions, e.g., the proposition that the input image depicts digit 2 or that the input prompt has a negative sentiment. This yields an interpretation of neurons, and groups thereof, that combines a logical and a Bayesian perspective, and accounts for polysemanticity (i.e., that a single neuron can figure in multiple "},"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":"2505.14424","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-05-20T14:32:03Z","cross_cats_sorted":[],"title_canon_sha256":"43f44e3ef97cfbcf6e87410d9eddd3e275551ab72e450694812355697c74de1b","abstract_canon_sha256":"1235303aba435131c7eab3c7563806f5b7d4458c1d986f6578f5ec5e88eadd4e"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:01.132267Z","signature_b64":"VjjVti9Mec/6zcgM3i+6EahvAUCuxvBZQWLz69sWOmxkr2GLNVE9ivfC3Qv9nWOCFH87ex2oUKgi5kLkF4UzCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"7c64d9e6f7b8f32b6d2552c2aea54efe37003274aabd29b951abc7e71c8ab583","last_reissued_at":"2026-07-05T11:06:01.131862Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:01.131862Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Explaining Neural Networks with Reasons","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Hannes Leitgeb, Levin Hornischer","submitted_at":"2025-05-20T14:32:03Z","abstract_excerpt":"We propose a new interpretability method for neural networks, which is based on a novel mathematico-philosophical theory of reasons. Our method computes a vector for each neuron, called its reasons vector. We then can compute how strongly this reasons vector speaks for various propositions, e.g., the proposition that the input image depicts digit 2 or that the input prompt has a negative sentiment. This yields an interpretation of neurons, and groups thereof, that combines a logical and a Bayesian perspective, and accounts for polysemanticity (i.e., that a single neuron can figure in multiple "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2505.14424","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/2505.14424/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":"2505.14424","created_at":"2026-07-05T11:06:01.131921+00:00"},{"alias_kind":"arxiv_version","alias_value":"2505.14424v1","created_at":"2026-07-05T11:06:01.131921+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2505.14424","created_at":"2026-07-05T11:06:01.131921+00:00"},{"alias_kind":"pith_short_12","alias_value":"PRSNTZXXXDZS","created_at":"2026-07-05T11:06:01.131921+00:00"},{"alias_kind":"pith_short_16","alias_value":"PRSNTZXXXDZSW3JF","created_at":"2026-07-05T11:06:01.131921+00:00"},{"alias_kind":"pith_short_8","alias_value":"PRSNTZXX","created_at":"2026-07-05T11:06:01.131921+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/PRSNTZXXXDZSW3JFKLBK5JKO7Y","json":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y.json","graph_json":"https://pith.science/api/pith-number/PRSNTZXXXDZSW3JFKLBK5JKO7Y/graph.json","events_json":"https://pith.science/api/pith-number/PRSNTZXXXDZSW3JFKLBK5JKO7Y/events.json","paper":"https://pith.science/paper/PRSNTZXX"},"agent_actions":{"view_html":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y","download_json":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y.json","view_paper":"https://pith.science/paper/PRSNTZXX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2505.14424&json=true","fetch_graph":"https://pith.science/api/pith-number/PRSNTZXXXDZSW3JFKLBK5JKO7Y/graph.json","fetch_events":"https://pith.science/api/pith-number/PRSNTZXXXDZSW3JFKLBK5JKO7Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y/action/storage_attestation","attest_author":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y/action/author_attestation","sign_citation":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y/action/citation_signature","submit_replication":"https://pith.science/pith/PRSNTZXXXDZSW3JFKLBK5JKO7Y/action/replication_record"}},"created_at":"2026-07-05T11:06:01.131921+00:00","updated_at":"2026-07-05T11:06:01.131921+00:00"}