{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:YHM6N2I35RYZ7IXRFMS3ZUSB3E","short_pith_number":"pith:YHM6N2I3","schema_version":"1.0","canonical_sha256":"c1d9e6e91bec719fa2f12b25bcd241d908c45b2acdd68ccef31c934eb6955ed0","source":{"kind":"arxiv","id":"2311.08968","version":2},"attestation_state":"computed","paper":{"title":"Identifying Linear Relational Concepts in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Anthony Hunter, David Chanin, Oana-Maria Camburu","submitted_at":"2023-11-15T14:01:41Z","abstract_excerpt":"Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations. However, for any human-interpretable concept, how can we find its direction in the latent space? We present a technique called linear relational concepts (LRC) for finding concept directions corresponding to human-interpretable concepts by first modeling the relation between subject and object as a linear relational embedding (LRE). We find that inverting the LRE and using earlier object layers results in a powerful technique for finding concept directions that outpe"},"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":"2311.08968","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2023-11-15T14:01:41Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"a782203e2410155fb7880f58a607857b601a030021d71a660766d27496aef641","abstract_canon_sha256":"2d1fd3e23e6103d2e4c012606525f886368c7f4e1e1f6b3fbc914832829c31a0"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:02:22.590835Z","signature_b64":"IUnlDSKAWQkn9xCPDbsvpG+H5Tz9m0caNJeISxfkeRYX5vw3zOsVqLEPJTxao8YbROCdEWIa2QwnnjzX53iYDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c1d9e6e91bec719fa2f12b25bcd241d908c45b2acdd68ccef31c934eb6955ed0","last_reissued_at":"2026-07-05T08:02:22.590279Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:02:22.590279Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Identifying Linear Relational Concepts in Large Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Anthony Hunter, David Chanin, Oana-Maria Camburu","submitted_at":"2023-11-15T14:01:41Z","abstract_excerpt":"Transformer language models (LMs) have been shown to represent concepts as directions in the latent space of hidden activations. However, for any human-interpretable concept, how can we find its direction in the latent space? We present a technique called linear relational concepts (LRC) for finding concept directions corresponding to human-interpretable concepts by first modeling the relation between subject and object as a linear relational embedding (LRE). We find that inverting the LRE and using earlier object layers results in a powerful technique for finding concept directions that outpe"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2311.08968","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/2311.08968/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":"2311.08968","created_at":"2026-07-05T08:02:22.590332+00:00"},{"alias_kind":"arxiv_version","alias_value":"2311.08968v2","created_at":"2026-07-05T08:02:22.590332+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2311.08968","created_at":"2026-07-05T08:02:22.590332+00:00"},{"alias_kind":"pith_short_12","alias_value":"YHM6N2I35RYZ","created_at":"2026-07-05T08:02:22.590332+00:00"},{"alias_kind":"pith_short_16","alias_value":"YHM6N2I35RYZ7IXR","created_at":"2026-07-05T08:02:22.590332+00:00"},{"alias_kind":"pith_short_8","alias_value":"YHM6N2I3","created_at":"2026-07-05T08:02:22.590332+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.27981","citing_title":"ToxiREX: A Dataset on Toxic REasoning in ConteXt","ref_index":258,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12299","citing_title":"GKnow: Measuring the Entanglement of Gender Bias and Factual Gender","ref_index":42,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E","json":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E.json","graph_json":"https://pith.science/api/pith-number/YHM6N2I35RYZ7IXRFMS3ZUSB3E/graph.json","events_json":"https://pith.science/api/pith-number/YHM6N2I35RYZ7IXRFMS3ZUSB3E/events.json","paper":"https://pith.science/paper/YHM6N2I3"},"agent_actions":{"view_html":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E","download_json":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E.json","view_paper":"https://pith.science/paper/YHM6N2I3","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2311.08968&json=true","fetch_graph":"https://pith.science/api/pith-number/YHM6N2I35RYZ7IXRFMS3ZUSB3E/graph.json","fetch_events":"https://pith.science/api/pith-number/YHM6N2I35RYZ7IXRFMS3ZUSB3E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E/action/storage_attestation","attest_author":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E/action/author_attestation","sign_citation":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E/action/citation_signature","submit_replication":"https://pith.science/pith/YHM6N2I35RYZ7IXRFMS3ZUSB3E/action/replication_record"}},"created_at":"2026-07-05T08:02:22.590332+00:00","updated_at":"2026-07-05T08:02:22.590332+00:00"}