{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:SUXZN6UGQQFPVEY5VWIBFRFFUU","short_pith_number":"pith:SUXZN6UG","schema_version":"1.0","canonical_sha256":"952f96fa86840afa931dad9012c4a5a51470c7a3c936a257ec7d9d642736e6f9","source":{"kind":"arxiv","id":"2402.17700","version":2},"attestation_state":"computed","paper":{"title":"RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Christopher Potts, Jing Huang, Mor Geva, Zhengxuan Wu","submitted_at":"2024-02-27T17:25:37Z","abstract_excerpt":"Individual neurons participate in the representation of multiple high-level concepts. To what extent can different interpretability methods successfully disentangle these roles? To help address this question, we introduce RAVEL (Resolving Attribute-Value Entanglements in Language Models), a dataset that enables tightly controlled, quantitative comparisons between a variety of existing interpretability methods. We use the resulting conceptual framework to define the new method of Multi-task Distributed Alignment Search (MDAS), which allows us to find distributed representations satisfying 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":"2402.17700","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-02-27T17:25:37Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"2438d7cf21176adec9d791d6429028b31e6e72efb89dd512d3d945d03b0e6573","abstract_canon_sha256":"39c096003002cd5e9532b00d02145c74ae2b5de2e77f2a62e8e690a67464edad"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:59:26.357732Z","signature_b64":"YqQfo3anrPIL7GPpwoUc8ECBFZ6Th8juT3TvVdjamP0lAFvJUV82ADcCG/sj0fV1X3/ZnOOzlOQR52Q/0UpSCw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"952f96fa86840afa931dad9012c4a5a51470c7a3c936a257ec7d9d642736e6f9","last_reissued_at":"2026-07-05T08:59:26.357237Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:59:26.357237Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"RAVEL: Evaluating Interpretability Methods on Disentangling Language Model Representations","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Atticus Geiger, Christopher Potts, Jing Huang, Mor Geva, Zhengxuan Wu","submitted_at":"2024-02-27T17:25:37Z","abstract_excerpt":"Individual neurons participate in the representation of multiple high-level concepts. To what extent can different interpretability methods successfully disentangle these roles? To help address this question, we introduce RAVEL (Resolving Attribute-Value Entanglements in Language Models), a dataset that enables tightly controlled, quantitative comparisons between a variety of existing interpretability methods. We use the resulting conceptual framework to define the new method of Multi-task Distributed Alignment Search (MDAS), which allows us to find distributed representations satisfying multi"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.17700","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/2402.17700/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":"2402.17700","created_at":"2026-07-05T08:59:26.357299+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.17700v2","created_at":"2026-07-05T08:59:26.357299+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.17700","created_at":"2026-07-05T08:59:26.357299+00:00"},{"alias_kind":"pith_short_12","alias_value":"SUXZN6UGQQFP","created_at":"2026-07-05T08:59:26.357299+00:00"},{"alias_kind":"pith_short_16","alias_value":"SUXZN6UGQQFPVEY5","created_at":"2026-07-05T08:59:26.357299+00:00"},{"alias_kind":"pith_short_8","alias_value":"SUXZN6UG","created_at":"2026-07-05T08:59:26.357299+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2507.05387","citing_title":"The Generalization Ridge: Information Flow in Natural Language Generation","ref_index":30,"is_internal_anchor":false},{"citing_arxiv_id":"2605.01381","citing_title":"A framework for analyzing concept representations in neural models","ref_index":102,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU","json":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU.json","graph_json":"https://pith.science/api/pith-number/SUXZN6UGQQFPVEY5VWIBFRFFUU/graph.json","events_json":"https://pith.science/api/pith-number/SUXZN6UGQQFPVEY5VWIBFRFFUU/events.json","paper":"https://pith.science/paper/SUXZN6UG"},"agent_actions":{"view_html":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU","download_json":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU.json","view_paper":"https://pith.science/paper/SUXZN6UG","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.17700&json=true","fetch_graph":"https://pith.science/api/pith-number/SUXZN6UGQQFPVEY5VWIBFRFFUU/graph.json","fetch_events":"https://pith.science/api/pith-number/SUXZN6UGQQFPVEY5VWIBFRFFUU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU/action/storage_attestation","attest_author":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU/action/author_attestation","sign_citation":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU/action/citation_signature","submit_replication":"https://pith.science/pith/SUXZN6UGQQFPVEY5VWIBFRFFUU/action/replication_record"}},"created_at":"2026-07-05T08:59:26.357299+00:00","updated_at":"2026-07-05T08:59:26.357299+00:00"}