{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:5ORDJC5NTSXAHUQYEE3F3WD7LP","short_pith_number":"pith:5ORDJC5N","schema_version":"1.0","canonical_sha256":"eba2348bad9cae03d21821365dd87f5bc2d284e6e7348b48e5027ce46a6f0101","source":{"kind":"arxiv","id":"2402.01702","version":1},"attestation_state":"computed","paper":{"title":"Fluent dreaming for language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Michael Sklar (1) ((1) Confirm Labs), T. Ben Thompson (1), Zygimantas Straznickas (1)","submitted_at":"2024-01-24T17:57:12Z","abstract_excerpt":"Feature visualization, also known as \"dreaming\", offers insights into vision models by optimizing the inputs to maximize a neuron's activation or other internal component. However, dreaming has not been successfully applied to language models because the input space is discrete. We extend Greedy Coordinate Gradient, a method from the language model adversarial attack literature, to design the Evolutionary Prompt Optimization (EPO) algorithm. EPO optimizes the input prompt to simultaneously maximize the Pareto frontier between a chosen internal feature and prompt fluency, enabling fluent dreami"},"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":"2402.01702","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2024-01-24T17:57:12Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"70fc6b85de45ddabadccc9a75aacd0f1092b9e05e93e8e3099a91eb1cf1c9e4d","abstract_canon_sha256":"86fbfa17d002df742fcdd7d94db5de252d643cb23e63588f5dc49b9b9689ce33"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:41:05.355133Z","signature_b64":"KCgZMB1MS86Dn/nfVaAe1HMm95yWiij5KFjWLk6Bp9oDFXUq6K8VEv3I4R3UfZ8YhWt70zRp77xL+rosk/w/Ag==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"eba2348bad9cae03d21821365dd87f5bc2d284e6e7348b48e5027ce46a6f0101","last_reissued_at":"2026-07-05T07:41:05.354613Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:41:05.354613Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Fluent dreaming for language models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CL","authors_text":"Michael Sklar (1) ((1) Confirm Labs), T. Ben Thompson (1), Zygimantas Straznickas (1)","submitted_at":"2024-01-24T17:57:12Z","abstract_excerpt":"Feature visualization, also known as \"dreaming\", offers insights into vision models by optimizing the inputs to maximize a neuron's activation or other internal component. However, dreaming has not been successfully applied to language models because the input space is discrete. We extend Greedy Coordinate Gradient, a method from the language model adversarial attack literature, to design the Evolutionary Prompt Optimization (EPO) algorithm. EPO optimizes the input prompt to simultaneously maximize the Pareto frontier between a chosen internal feature and prompt fluency, enabling fluent dreami"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.01702","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/2402.01702/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":"2402.01702","created_at":"2026-07-05T07:41:05.354675+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.01702v1","created_at":"2026-07-05T07:41:05.354675+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.01702","created_at":"2026-07-05T07:41:05.354675+00:00"},{"alias_kind":"pith_short_12","alias_value":"5ORDJC5NTSXA","created_at":"2026-07-05T07:41:05.354675+00:00"},{"alias_kind":"pith_short_16","alias_value":"5ORDJC5NTSXAHUQY","created_at":"2026-07-05T07:41:05.354675+00:00"},{"alias_kind":"pith_short_8","alias_value":"5ORDJC5N","created_at":"2026-07-05T07:41:05.354675+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP","json":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP.json","graph_json":"https://pith.science/api/pith-number/5ORDJC5NTSXAHUQYEE3F3WD7LP/graph.json","events_json":"https://pith.science/api/pith-number/5ORDJC5NTSXAHUQYEE3F3WD7LP/events.json","paper":"https://pith.science/paper/5ORDJC5N"},"agent_actions":{"view_html":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP","download_json":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP.json","view_paper":"https://pith.science/paper/5ORDJC5N","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.01702&json=true","fetch_graph":"https://pith.science/api/pith-number/5ORDJC5NTSXAHUQYEE3F3WD7LP/graph.json","fetch_events":"https://pith.science/api/pith-number/5ORDJC5NTSXAHUQYEE3F3WD7LP/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP/action/timestamp_anchor","attest_storage":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP/action/storage_attestation","attest_author":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP/action/author_attestation","sign_citation":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP/action/citation_signature","submit_replication":"https://pith.science/pith/5ORDJC5NTSXAHUQYEE3F3WD7LP/action/replication_record"}},"created_at":"2026-07-05T07:41:05.354675+00:00","updated_at":"2026-07-05T07:41:05.354675+00:00"}