{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:TNE5F3T4PXLXLV6GZMYM5SYHVS","short_pith_number":"pith:TNE5F3T4","schema_version":"1.0","canonical_sha256":"9b49d2ee7c7dd775d7c6cb30cecb07ac8bd6c2b398b84dff32122ce6c49d6a3b","source":{"kind":"arxiv","id":"2606.00356","version":1},"attestation_state":"computed","paper":{"title":"How Far Do Auto-Interpretation Labels Generalize: A Controlled Study Across Languages, Scripts, and Rewordings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Sripad Karne","submitted_at":"2026-05-29T20:59:28Z","abstract_excerpt":"Sparse autoencoder (SAE) features are increasingly used to interpret language models, with auto-generated natural-language labels serving as the primary interface for understanding what each feature represents. We ask whether these labels generalize: does a feature labeled for a concept actually track that concept across languages and scripts? Using Serbian digraphia as a controlled testbed -- the same language written in both Latin and Cyrillic via deterministic transliteration -- we first find that SAE feature sets activated by the same content in different languages, scripts, and wordings s"},"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":"2606.00356","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2026-05-29T20:59:28Z","cross_cats_sorted":[],"title_canon_sha256":"c76cd184ede58f49da64d34a24c8c36832a3df4e2686c9bdbd8c9448c3408d1d","abstract_canon_sha256":"a47e490c28b6b1d50607b2878a32a04492a51ef26c5ededcd773d9eb4e588bfb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-02T01:03:52.512655Z","signature_b64":"8ayOE1N6UVVM4iqBHlDuROOeBoXwX2k4Fim74fA/rR6NyqBbZp0YAcy5DYujzMBYFCmDhND7WSZe0rcmwPJ1Bg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"9b49d2ee7c7dd775d7c6cb30cecb07ac8bd6c2b398b84dff32122ce6c49d6a3b","last_reissued_at":"2026-06-02T01:03:52.512271Z","signature_status":"signed_v1","first_computed_at":"2026-06-02T01:03:52.512271Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"How Far Do Auto-Interpretation Labels Generalize: A Controlled Study Across Languages, Scripts, and Rewordings","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.CL","authors_text":"Sripad Karne","submitted_at":"2026-05-29T20:59:28Z","abstract_excerpt":"Sparse autoencoder (SAE) features are increasingly used to interpret language models, with auto-generated natural-language labels serving as the primary interface for understanding what each feature represents. We ask whether these labels generalize: does a feature labeled for a concept actually track that concept across languages and scripts? Using Serbian digraphia as a controlled testbed -- the same language written in both Latin and Cyrillic via deterministic transliteration -- we first find that SAE feature sets activated by the same content in different languages, scripts, and wordings s"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.00356","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/2606.00356/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":"2606.00356","created_at":"2026-06-02T01:03:52.512325+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.00356v1","created_at":"2026-06-02T01:03:52.512325+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.00356","created_at":"2026-06-02T01:03:52.512325+00:00"},{"alias_kind":"pith_short_12","alias_value":"TNE5F3T4PXLX","created_at":"2026-06-02T01:03:52.512325+00:00"},{"alias_kind":"pith_short_16","alias_value":"TNE5F3T4PXLXLV6G","created_at":"2026-06-02T01:03:52.512325+00:00"},{"alias_kind":"pith_short_8","alias_value":"TNE5F3T4","created_at":"2026-06-02T01:03:52.512325+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/TNE5F3T4PXLXLV6GZMYM5SYHVS","json":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS.json","graph_json":"https://pith.science/api/pith-number/TNE5F3T4PXLXLV6GZMYM5SYHVS/graph.json","events_json":"https://pith.science/api/pith-number/TNE5F3T4PXLXLV6GZMYM5SYHVS/events.json","paper":"https://pith.science/paper/TNE5F3T4"},"agent_actions":{"view_html":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS","download_json":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS.json","view_paper":"https://pith.science/paper/TNE5F3T4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.00356&json=true","fetch_graph":"https://pith.science/api/pith-number/TNE5F3T4PXLXLV6GZMYM5SYHVS/graph.json","fetch_events":"https://pith.science/api/pith-number/TNE5F3T4PXLXLV6GZMYM5SYHVS/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS/action/timestamp_anchor","attest_storage":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS/action/storage_attestation","attest_author":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS/action/author_attestation","sign_citation":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS/action/citation_signature","submit_replication":"https://pith.science/pith/TNE5F3T4PXLXLV6GZMYM5SYHVS/action/replication_record"}},"created_at":"2026-06-02T01:03:52.512325+00:00","updated_at":"2026-06-02T01:03:52.512325+00:00"}