{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:JIHVYSHDYO2CNSQGW3WTOP44WO","short_pith_number":"pith:JIHVYSHD","schema_version":"1.0","canonical_sha256":"4a0f5c48e3c3b426ca06b6ed373f9cb3b3043e0ccc8a1029d0292eb8870cd681","source":{"kind":"arxiv","id":"2406.01506","version":3},"attestation_state":"computed","paper":{"title":"The Geometry of Categorical and Hierarchical Concepts in Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Kiho Park, Victor Veitch, Yibo Jiang, Yo Joong Choe","submitted_at":"2024-06-03T16:34:01Z","abstract_excerpt":"The linear representation hypothesis is the informal idea that semantic concepts are encoded as linear directions in the representation spaces of large language models (LLMs). Previous work has shown how to make this notion precise for representing binary concepts that have natural contrasts (e.g., {male, female}) as directions in representation space. However, many natural concepts do not have natural contrasts (e.g., whether the output is about an animal). In this work, we show how to extend the formalization of the linear representation hypothesis to represent features (e.g., is_animal) as "},"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":"2406.01506","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","primary_cat":"cs.CL","submitted_at":"2024-06-03T16:34:01Z","cross_cats_sorted":["cs.AI","cs.LG","stat.ML"],"title_canon_sha256":"8616a674f7fe7ffc57362fe989500db6acc6bb2794d4d0797ab591699289ea2b","abstract_canon_sha256":"999efc8566365635d72cd13f3aae2d38bcd703ca6c4802218afc817008d77feb"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:15:43.900374Z","signature_b64":"9AM8Q8XwC7Ijt/rLLaoGLR3W366WiXwx2UufvAnLqR8qkPSehtJXl9oJ9aFoqt0Wa3IwdejH30eruLjGpCp+CA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4a0f5c48e3c3b426ca06b6ed373f9cb3b3043e0ccc8a1029d0292eb8870cd681","last_reissued_at":"2026-07-05T10:15:43.899864Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:15:43.899864Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"The Geometry of Categorical and Hierarchical Concepts in Large Language Models","license":"http://creativecommons.org/licenses/by-nc-sa/4.0/","headline":"","cross_cats":["cs.AI","cs.LG","stat.ML"],"primary_cat":"cs.CL","authors_text":"Kiho Park, Victor Veitch, Yibo Jiang, Yo Joong Choe","submitted_at":"2024-06-03T16:34:01Z","abstract_excerpt":"The linear representation hypothesis is the informal idea that semantic concepts are encoded as linear directions in the representation spaces of large language models (LLMs). Previous work has shown how to make this notion precise for representing binary concepts that have natural contrasts (e.g., {male, female}) as directions in representation space. However, many natural concepts do not have natural contrasts (e.g., whether the output is about an animal). In this work, we show how to extend the formalization of the linear representation hypothesis to represent features (e.g., is_animal) as "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2406.01506","kind":"arxiv","version":3},"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/2406.01506/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":"2406.01506","created_at":"2026-07-05T10:15:43.899917+00:00"},{"alias_kind":"arxiv_version","alias_value":"2406.01506v3","created_at":"2026-07-05T10:15:43.899917+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2406.01506","created_at":"2026-07-05T10:15:43.899917+00:00"},{"alias_kind":"pith_short_12","alias_value":"JIHVYSHDYO2C","created_at":"2026-07-05T10:15:43.899917+00:00"},{"alias_kind":"pith_short_16","alias_value":"JIHVYSHDYO2CNSQG","created_at":"2026-07-05T10:15:43.899917+00:00"},{"alias_kind":"pith_short_8","alias_value":"JIHVYSHD","created_at":"2026-07-05T10:15:43.899917+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":21,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.07047","citing_title":"Riemannian Geometry for Pre-trained Language Model Embeddings","ref_index":41,"is_internal_anchor":true},{"citing_arxiv_id":"2606.25234","citing_title":"Structuring Sparsity: Block-Sparse Featurizers Capture Visual Concept Manifolds","ref_index":95,"is_internal_anchor":false},{"citing_arxiv_id":"2606.18284","citing_title":"Breaking the Solver Bottleneck: Training Task Generators at the Learnable Frontier","ref_index":122,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09607","citing_title":"Closure-Validated Circuit Discovery in Attention Heads: Co-activation Proposes, Ablation Disposes","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2606.09894","citing_title":"A Navigable Manifold of Hypothesized Consciousness-Spectrum States in Language Model Representations","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2606.15054","citing_title":"Size Doesn't Matter: Cosine-Scored Sparse Autoencoders","ref_index":6,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06333","citing_title":"Subspace-Aware Sparse Autoencoders for Effective Mechanistic Interpretability","ref_index":24,"is_internal_anchor":false},{"citing_arxiv_id":"2605.23033","citing_title":"Uncovering the Latent Potential of Deep Intermediate Representations","ref_index":4,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12765","citing_title":"Inference-Time Machine Unlearning via Gated Activation Redirection","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2605.17084","citing_title":"Scale Determines Whether Language Models Organize Representation Geometry for Prediction","ref_index":7,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12765","citing_title":"Inference-Time Machine Unlearning via Gated Activation Redirection","ref_index":40,"is_internal_anchor":false},{"citing_arxiv_id":"2604.27169","citing_title":"Semantic Structure of Feature Space in Large Language Models","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2604.26341","citing_title":"SpatialFusion: Endowing Unified Image Generation with Intrinsic 3D Geometric Awareness","ref_index":36,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07922","citing_title":"Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06196","citing_title":"The Granularity Axis: A Micro-to-Macro Latent Direction for Social Roles in Language Models","ref_index":41,"is_internal_anchor":false},{"citing_arxiv_id":"2605.05443","citing_title":"SLAM: Structural Linguistic Activation Marking for Language Models","ref_index":20,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07148","citing_title":"Uncovering and Shaping the Latent Representation of 3D Scene Topology in Vision-Language Models","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.07922","citing_title":"Tree SAE: Learning Hierarchical Feature Structures in Sparse Autoencoders","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.00847","citing_title":"H-Probes: Extracting Hierarchical Structures From Latent Representations of Language Models","ref_index":16,"is_internal_anchor":false},{"citing_arxiv_id":"2604.14363","citing_title":"The Cost of Language: Centroid Erasure Exposes and Exploits Modal Competition in Multimodal Language Models","ref_index":15,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01381","citing_title":"A framework for analyzing concept representations in neural models","ref_index":148,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO","json":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO.json","graph_json":"https://pith.science/api/pith-number/JIHVYSHDYO2CNSQGW3WTOP44WO/graph.json","events_json":"https://pith.science/api/pith-number/JIHVYSHDYO2CNSQGW3WTOP44WO/events.json","paper":"https://pith.science/paper/JIHVYSHD"},"agent_actions":{"view_html":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO","download_json":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO.json","view_paper":"https://pith.science/paper/JIHVYSHD","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2406.01506&json=true","fetch_graph":"https://pith.science/api/pith-number/JIHVYSHDYO2CNSQGW3WTOP44WO/graph.json","fetch_events":"https://pith.science/api/pith-number/JIHVYSHDYO2CNSQGW3WTOP44WO/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO/action/storage_attestation","attest_author":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO/action/author_attestation","sign_citation":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO/action/citation_signature","submit_replication":"https://pith.science/pith/JIHVYSHDYO2CNSQGW3WTOP44WO/action/replication_record"}},"created_at":"2026-07-05T10:15:43.899917+00:00","updated_at":"2026-07-05T10:15:43.899917+00:00"}