{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:LILRSDWZGGYJAKF3T52IGRIQIV","short_pith_number":"pith:LILRSDWZ","schema_version":"1.0","canonical_sha256":"5a17190ed931b09028bb9f7483451045596c328575f93ec19ded832643264e29","source":{"kind":"arxiv","id":"2305.09863","version":2},"attestation_state":"computed","paper":{"title":"Explaining black box text modules in natural language with language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG","q-bio.NC"],"primary_cat":"cs.AI","authors_text":"Alexander G. Huth, Aliyah R. Hsu, Bin Yu, Chandan Singh, Jianfeng Gao, Richard Antonello, Shailee Jain","submitted_at":"2023-05-17T00:29:18Z","abstract_excerpt":"Large language models (LLMs) have demonstrated remarkable prediction performance for a growing array of tasks. However, their rapid proliferation and increasing opaqueness have created a growing need for interpretability. Here, we ask whether we can automatically obtain natural language explanations for black box text modules. A \"text module\" is any function that maps text to a scalar continuous value, such as a submodule within an LLM or a fitted model of a brain region. \"Black box\" indicates that we only have access to the module's inputs/outputs.\n  We introduce Summarize and Score (SASC), a"},"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":"2305.09863","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.AI","submitted_at":"2023-05-17T00:29:18Z","cross_cats_sorted":["cs.CL","cs.LG","q-bio.NC"],"title_canon_sha256":"0e008d932fbecdd4316d5e25f49bb641fdee414757814c8454f53a1f5712e06d","abstract_canon_sha256":"eb259143bcd37fd775e75fa50818c43ec910d06ebf94022321564e39db12edaf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:12:50.013613Z","signature_b64":"PwTj5Vw1rrBLJyQlNqrXAEUqnf2OVqRCkJHx6rr6uVad9mjlJZ6yMIVbA++D5XGljVaEv3DNgMss3lmyfj7sCQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"5a17190ed931b09028bb9f7483451045596c328575f93ec19ded832643264e29","last_reissued_at":"2026-07-05T07:12:50.013185Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:12:50.013185Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Explaining black box text modules in natural language with language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL","cs.LG","q-bio.NC"],"primary_cat":"cs.AI","authors_text":"Alexander G. Huth, Aliyah R. Hsu, Bin Yu, Chandan Singh, Jianfeng Gao, Richard Antonello, Shailee Jain","submitted_at":"2023-05-17T00:29:18Z","abstract_excerpt":"Large language models (LLMs) have demonstrated remarkable prediction performance for a growing array of tasks. However, their rapid proliferation and increasing opaqueness have created a growing need for interpretability. Here, we ask whether we can automatically obtain natural language explanations for black box text modules. A \"text module\" is any function that maps text to a scalar continuous value, such as a submodule within an LLM or a fitted model of a brain region. \"Black box\" indicates that we only have access to the module's inputs/outputs.\n  We introduce Summarize and Score (SASC), a"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2305.09863","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/2305.09863/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":"2305.09863","created_at":"2026-07-05T07:12:50.013241+00:00"},{"alias_kind":"arxiv_version","alias_value":"2305.09863v2","created_at":"2026-07-05T07:12:50.013241+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2305.09863","created_at":"2026-07-05T07:12:50.013241+00:00"},{"alias_kind":"pith_short_12","alias_value":"LILRSDWZGGYJ","created_at":"2026-07-05T07:12:50.013241+00:00"},{"alias_kind":"pith_short_16","alias_value":"LILRSDWZGGYJAKF3","created_at":"2026-07-05T07:12:50.013241+00:00"},{"alias_kind":"pith_short_8","alias_value":"LILRSDWZ","created_at":"2026-07-05T07:12:50.013241+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.07007","citing_title":"A Geometric View for Understanding Concept Learning and Neuron Interpretation in Sparse Autoencoders","ref_index":31,"is_internal_anchor":false},{"citing_arxiv_id":"2509.13316","citing_title":"Do Activation Verbalization Methods Convey Privileged Information?","ref_index":48,"is_internal_anchor":false},{"citing_arxiv_id":"2605.03808","citing_title":"Agentic-imodels: Evolving agentic interpretability tools via autoresearch","ref_index":45,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV","json":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV.json","graph_json":"https://pith.science/api/pith-number/LILRSDWZGGYJAKF3T52IGRIQIV/graph.json","events_json":"https://pith.science/api/pith-number/LILRSDWZGGYJAKF3T52IGRIQIV/events.json","paper":"https://pith.science/paper/LILRSDWZ"},"agent_actions":{"view_html":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV","download_json":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV.json","view_paper":"https://pith.science/paper/LILRSDWZ","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2305.09863&json=true","fetch_graph":"https://pith.science/api/pith-number/LILRSDWZGGYJAKF3T52IGRIQIV/graph.json","fetch_events":"https://pith.science/api/pith-number/LILRSDWZGGYJAKF3T52IGRIQIV/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV/action/timestamp_anchor","attest_storage":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV/action/storage_attestation","attest_author":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV/action/author_attestation","sign_citation":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV/action/citation_signature","submit_replication":"https://pith.science/pith/LILRSDWZGGYJAKF3T52IGRIQIV/action/replication_record"}},"created_at":"2026-07-05T07:12:50.013241+00:00","updated_at":"2026-07-05T07:12:50.013241+00:00"}