{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:FUM6QLJX2NOOWPRC4L37XUE3RY","short_pith_number":"pith:FUM6QLJX","schema_version":"1.0","canonical_sha256":"2d19e82d37d35ceb3e22e2f7fbd09b8e2aa315346592747d0e0f1569940a5a54","source":{"kind":"arxiv","id":"2501.08319","version":2},"attestation_state":"computed","paper":{"title":"Enhancing Automated Interpretability with Output-Centric Feature Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Chen Agassy, Mor Geva, Roy Mayan, Yoav Gur-Arieh","submitted_at":"2025-01-14T18:53:00Z","abstract_excerpt":"Automated interpretability pipelines generate natural language descriptions for the concepts represented by features in large language models (LLMs), such as plants or the first word in a sentence. These descriptions are derived using inputs that activate the feature, which may be a dimension or a direction in the model's representation space. However, identifying activating inputs is costly, and the mechanistic role of a feature in model behavior is determined both by how inputs cause a feature to activate and by how feature activation affects outputs. Using steering evaluations, we reveal th"},"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":"2501.08319","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2025-01-14T18:53:00Z","cross_cats_sorted":[],"title_canon_sha256":"7cefc6dee2a4ae8f7daf899180dee7ccef20cd813f471f592a9f31e4b679769b","abstract_canon_sha256":"37ac9065a68be5aa2301227117527340d781c52b964d73e9cb194f5b3fa25ff2"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:11:39.716173Z","signature_b64":"44SiAz6yidn6yek3gJjEKKaNFj/9mesIIfeQE/JPdrWzJNOnjOsPtzFBTNZRSrnjtWrgWn2xdV/LJamDFr5DCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"2d19e82d37d35ceb3e22e2f7fbd09b8e2aa315346592747d0e0f1569940a5a54","last_reissued_at":"2026-07-05T11:11:39.715624Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:11:39.715624Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Enhancing Automated Interpretability with Output-Centric Feature Descriptions","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Chen Agassy, Mor Geva, Roy Mayan, Yoav Gur-Arieh","submitted_at":"2025-01-14T18:53:00Z","abstract_excerpt":"Automated interpretability pipelines generate natural language descriptions for the concepts represented by features in large language models (LLMs), such as plants or the first word in a sentence. These descriptions are derived using inputs that activate the feature, which may be a dimension or a direction in the model's representation space. However, identifying activating inputs is costly, and the mechanistic role of a feature in model behavior is determined both by how inputs cause a feature to activate and by how feature activation affects outputs. Using steering evaluations, we reveal th"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2501.08319","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/2501.08319/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":"2501.08319","created_at":"2026-07-05T11:11:39.715689+00:00"},{"alias_kind":"arxiv_version","alias_value":"2501.08319v2","created_at":"2026-07-05T11:11:39.715689+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2501.08319","created_at":"2026-07-05T11:11:39.715689+00:00"},{"alias_kind":"pith_short_12","alias_value":"FUM6QLJX2NOO","created_at":"2026-07-05T11:11:39.715689+00:00"},{"alias_kind":"pith_short_16","alias_value":"FUM6QLJX2NOOWPRC","created_at":"2026-07-05T11:11:39.715689+00:00"},{"alias_kind":"pith_short_8","alias_value":"FUM6QLJX","created_at":"2026-07-05T11:11:39.715689+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27237","citing_title":"LMs as Task-Specific Knowledge Bases: An Interpretability Analysis","ref_index":65,"is_internal_anchor":false},{"citing_arxiv_id":"2601.14004","citing_title":"Locate, Steer, and Improve: A Practical Survey of Actionable Mechanistic Interpretability in Large Language Models","ref_index":96,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12299","citing_title":"GKnow: Measuring the Entanglement of Gender Bias and Factual Gender","ref_index":52,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY","json":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY.json","graph_json":"https://pith.science/api/pith-number/FUM6QLJX2NOOWPRC4L37XUE3RY/graph.json","events_json":"https://pith.science/api/pith-number/FUM6QLJX2NOOWPRC4L37XUE3RY/events.json","paper":"https://pith.science/paper/FUM6QLJX"},"agent_actions":{"view_html":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY","download_json":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY.json","view_paper":"https://pith.science/paper/FUM6QLJX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2501.08319&json=true","fetch_graph":"https://pith.science/api/pith-number/FUM6QLJX2NOOWPRC4L37XUE3RY/graph.json","fetch_events":"https://pith.science/api/pith-number/FUM6QLJX2NOOWPRC4L37XUE3RY/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY/action/timestamp_anchor","attest_storage":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY/action/storage_attestation","attest_author":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY/action/author_attestation","sign_citation":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY/action/citation_signature","submit_replication":"https://pith.science/pith/FUM6QLJX2NOOWPRC4L37XUE3RY/action/replication_record"}},"created_at":"2026-07-05T11:11:39.715689+00:00","updated_at":"2026-07-05T11:11:39.715689+00:00"}