{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:YFPJSDAXKO2WTKJLSTLOYJGH2Q","short_pith_number":"pith:YFPJSDAX","schema_version":"1.0","canonical_sha256":"c15e990c1753b569a92b94d6ec24c7d433330c3b9713618e592577e870740807","source":{"kind":"arxiv","id":"2506.12152","version":1},"attestation_state":"computed","paper":{"title":"Because we have LLMs, we Can and Should Pursue Agentic Interpretability","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Been Kim, John Hewitt, Neel Nanda, Noah Fiedel, Oyvind Tafjord","submitted_at":"2025-06-13T18:13:58Z","abstract_excerpt":"The era of Large Language Models (LLMs) presents a new opportunity for interpretability--agentic interpretability: a multi-turn conversation with an LLM wherein the LLM proactively assists human understanding by developing and leveraging a mental model of the user, which in turn enables humans to develop better mental models of the LLM. Such conversation is a new capability that traditional `inspective' interpretability methods (opening the black-box) do not use. Having a language model that aims to teach and explain--beyond just knowing how to talk--is similar to a teacher whose goal is to te"},"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":"2506.12152","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.AI","submitted_at":"2025-06-13T18:13:58Z","cross_cats_sorted":[],"title_canon_sha256":"5475c2691d851023d95d3b863d6aae4c72ad83664d3d7106d4d93185d214e9f2","abstract_canon_sha256":"c07c0ceecfffb8f619109fe5689670ee510b74ef89b6c10faf4a80309c06f745"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:21:22.094128Z","signature_b64":"jdmPYUCUszchE89p3KMINGBJI86PQmgMsXbAgrYd9ImatBe1p2hCNpXIHsOS8sFWATUqNRzkcOy+vB/4vd1tDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"c15e990c1753b569a92b94d6ec24c7d433330c3b9713618e592577e870740807","last_reissued_at":"2026-07-05T11:21:22.093532Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:21:22.093532Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Because we have LLMs, we Can and Should Pursue Agentic Interpretability","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":[],"primary_cat":"cs.AI","authors_text":"Been Kim, John Hewitt, Neel Nanda, Noah Fiedel, Oyvind Tafjord","submitted_at":"2025-06-13T18:13:58Z","abstract_excerpt":"The era of Large Language Models (LLMs) presents a new opportunity for interpretability--agentic interpretability: a multi-turn conversation with an LLM wherein the LLM proactively assists human understanding by developing and leveraging a mental model of the user, which in turn enables humans to develop better mental models of the LLM. Such conversation is a new capability that traditional `inspective' interpretability methods (opening the black-box) do not use. Having a language model that aims to teach and explain--beyond just knowing how to talk--is similar to a teacher whose goal is to te"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2506.12152","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/2506.12152/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":"2506.12152","created_at":"2026-07-05T11:21:22.093614+00:00"},{"alias_kind":"arxiv_version","alias_value":"2506.12152v1","created_at":"2026-07-05T11:21:22.093614+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2506.12152","created_at":"2026-07-05T11:21:22.093614+00:00"},{"alias_kind":"pith_short_12","alias_value":"YFPJSDAXKO2W","created_at":"2026-07-05T11:21:22.093614+00:00"},{"alias_kind":"pith_short_16","alias_value":"YFPJSDAXKO2WTKJL","created_at":"2026-07-05T11:21:22.093614+00:00"},{"alias_kind":"pith_short_8","alias_value":"YFPJSDAX","created_at":"2026-07-05T11:21:22.093614+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":4,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.19559","citing_title":"Uncertainty Decomposition for Clarification Seeking in LLM Agents","ref_index":14,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07007","citing_title":"A Geometric View for Understanding Concept Learning and Neuron Interpretation in Sparse Autoencoders","ref_index":55,"is_internal_anchor":false},{"citing_arxiv_id":"2606.27510","citing_title":"The Curse of Multiple Mediators: Hidden Interaction Effects in Activation Patching","ref_index":8,"is_internal_anchor":false},{"citing_arxiv_id":"2505.15436","citing_title":"Adaptive Chain-of-Focus Reasoning via Dynamic Visual Search and Zooming for Efficient VLMs","ref_index":29,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q","json":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q.json","graph_json":"https://pith.science/api/pith-number/YFPJSDAXKO2WTKJLSTLOYJGH2Q/graph.json","events_json":"https://pith.science/api/pith-number/YFPJSDAXKO2WTKJLSTLOYJGH2Q/events.json","paper":"https://pith.science/paper/YFPJSDAX"},"agent_actions":{"view_html":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q","download_json":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q.json","view_paper":"https://pith.science/paper/YFPJSDAX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2506.12152&json=true","fetch_graph":"https://pith.science/api/pith-number/YFPJSDAXKO2WTKJLSTLOYJGH2Q/graph.json","fetch_events":"https://pith.science/api/pith-number/YFPJSDAXKO2WTKJLSTLOYJGH2Q/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q/action/timestamp_anchor","attest_storage":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q/action/storage_attestation","attest_author":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q/action/author_attestation","sign_citation":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q/action/citation_signature","submit_replication":"https://pith.science/pith/YFPJSDAXKO2WTKJLSTLOYJGH2Q/action/replication_record"}},"created_at":"2026-07-05T11:21:22.093614+00:00","updated_at":"2026-07-05T11:21:22.093614+00:00"}