{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:64J3OPNBEVFDOPZQI3SYYJNCD3","short_pith_number":"pith:64J3OPNB","schema_version":"1.0","canonical_sha256":"f713b73da1254a373f3046e58c25a21ec2b5f43538195b51ed49d64d750eebdd","source":{"kind":"arxiv","id":"2410.06981","version":4},"attestation_state":"computed","paper":{"title":"Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Ashkan Khakzar, Austin Meek, David Krueger, Fazl Barez, Michael Lan, Philip Torr","submitted_at":"2024-10-09T15:18:57Z","abstract_excerpt":"The Universality Hypothesis in large language models (LLMs) claims that different models converge towards similar concept representations in their latent spaces. Providing evidence for this hypothesis would enable researchers to exploit universal properties, facilitating the generalization of mechanistic interpretability techniques across models. Previous works studied if LLMs learned the same features, which are internal representations that activate on specific concepts. Since comparing features across LLMs is challenging due to polysemanticity, in which LLM neurons often correspond to multi"},"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":"2410.06981","kind":"arxiv","version":4},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2024-10-09T15:18:57Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"84f40a433b8c4e6a3a517e38d82a38d6fff5fc345c939ae8ca82de7081eee474","abstract_canon_sha256":"9d0611d37345f59dfa2d6c0e30a8afc0c6319d1536082559606171ec020f23bf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:06:23.949550Z","signature_b64":"pGqZqG6c4rutSri98OCpYqaIr81Y7b99JdiHCVjM1esNhNqljsLfns4OOYMkSrjIxwypGv0g4u5ENSPsuwDBBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f713b73da1254a373f3046e58c25a21ec2b5f43538195b51ed49d64d750eebdd","last_reissued_at":"2026-07-05T11:06:23.949039Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:06:23.949039Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Quantifying Feature Space Universality Across Large Language Models via Sparse Autoencoders","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.LG","authors_text":"Ashkan Khakzar, Austin Meek, David Krueger, Fazl Barez, Michael Lan, Philip Torr","submitted_at":"2024-10-09T15:18:57Z","abstract_excerpt":"The Universality Hypothesis in large language models (LLMs) claims that different models converge towards similar concept representations in their latent spaces. Providing evidence for this hypothesis would enable researchers to exploit universal properties, facilitating the generalization of mechanistic interpretability techniques across models. Previous works studied if LLMs learned the same features, which are internal representations that activate on specific concepts. Since comparing features across LLMs is challenging due to polysemanticity, in which LLM neurons often correspond to multi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2410.06981","kind":"arxiv","version":4},"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/2410.06981/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":"2410.06981","created_at":"2026-07-05T11:06:23.949091+00:00"},{"alias_kind":"arxiv_version","alias_value":"2410.06981v4","created_at":"2026-07-05T11:06:23.949091+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2410.06981","created_at":"2026-07-05T11:06:23.949091+00:00"},{"alias_kind":"pith_short_12","alias_value":"64J3OPNBEVFD","created_at":"2026-07-05T11:06:23.949091+00:00"},{"alias_kind":"pith_short_16","alias_value":"64J3OPNBEVFDOPZQ","created_at":"2026-07-05T11:06:23.949091+00:00"},{"alias_kind":"pith_short_8","alias_value":"64J3OPNB","created_at":"2026-07-05T11:06:23.949091+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":10,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.24577","citing_title":"Polymorphism Is Rotation: Operational Mechanistic Interpretability from a Two-Layer Transformer to Pythia-70m","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2605.22202","citing_title":"Structure Retention in Embedding Spaces as a Predictor of Benchmark Performance","ref_index":102,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":25,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19765","citing_title":"Do Hallucination Neurons Generalize? Evidence from Cross-Domain Transfer in LLMs","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12770","citing_title":"WriteSAE: Sparse Autoencoders for Recurrent State","ref_index":80,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01609","citing_title":"Concepts Whisper While Syntax Shouts: Spectral Anti-Concentration and the Dual Geometry of Transformer Representations","ref_index":11,"is_internal_anchor":false},{"citing_arxiv_id":"2604.19260","citing_title":"Understanding the Mechanism of Altruism in Large Language Models","ref_index":243,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05508","citing_title":"Rigorous Interpretation Is a Form of Evaluation","ref_index":118,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3","json":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3.json","graph_json":"https://pith.science/api/pith-number/64J3OPNBEVFDOPZQI3SYYJNCD3/graph.json","events_json":"https://pith.science/api/pith-number/64J3OPNBEVFDOPZQI3SYYJNCD3/events.json","paper":"https://pith.science/paper/64J3OPNB"},"agent_actions":{"view_html":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3","download_json":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3.json","view_paper":"https://pith.science/paper/64J3OPNB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2410.06981&json=true","fetch_graph":"https://pith.science/api/pith-number/64J3OPNBEVFDOPZQI3SYYJNCD3/graph.json","fetch_events":"https://pith.science/api/pith-number/64J3OPNBEVFDOPZQI3SYYJNCD3/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3/action/timestamp_anchor","attest_storage":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3/action/storage_attestation","attest_author":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3/action/author_attestation","sign_citation":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3/action/citation_signature","submit_replication":"https://pith.science/pith/64J3OPNBEVFDOPZQI3SYYJNCD3/action/replication_record"}},"created_at":"2026-07-05T11:06:23.949091+00:00","updated_at":"2026-07-05T11:06:23.949091+00:00"}