{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:N4VWJGE4AZSS4BAL2ZPV7A6CVZ","short_pith_number":"pith:N4VWJGE4","schema_version":"1.0","canonical_sha256":"6f2b64989c06652e040bd65f5f83c2ae682b625ffaa7731496daacaa6305beae","source":{"kind":"arxiv","id":"2402.08756","version":1},"attestation_state":"computed","paper":{"title":"Learning How To Ask: Cycle-Consistency Refines Prompts in Multimodal Foundation Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Gayathri Mahalingam, Jianzhe Lin, Jie Zhao, Maurice Diesendruck, Mingyang Xu, Shima Imani","submitted_at":"2024-02-13T19:49:17Z","abstract_excerpt":"When LLMs perform zero-shot inference, they typically use a prompt with a task specification, and generate a completion. However, there is no work to explore the possibility of the reverse - going from completion to task specification. In this paper, we employ both directions to perform cycle-supervised learning entirely in-context. Our goal is to create a forward map f : X -> Y (e.g. image -> generated caption), coupled with a backward map g : Y -> X (e.g. caption -> generated image) to construct a cycle-consistency \"loss\" (formulated as an update to the prompt) to enforce g(f(X)) ~= X. The t"},"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":"2402.08756","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CL","submitted_at":"2024-02-13T19:49:17Z","cross_cats_sorted":["cs.CV"],"title_canon_sha256":"04589798f0582c9303eceb235d88341a7db1de8daf55264ee6574bee0f5a9899","abstract_canon_sha256":"103272d75914a61ad591663b2c0db50616f1185264b61b4275238bea42deac8d"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T07:45:02.007188Z","signature_b64":"T2lAVlt8gom7mZIHB1yq2xeiAORfJTDlNYh0L1nRACBI3AjEL2BgiKH5SlVauos3Ltp5W28kPWB+SEak5SkLAg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"6f2b64989c06652e040bd65f5f83c2ae682b625ffaa7731496daacaa6305beae","last_reissued_at":"2026-07-05T07:45:02.006750Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T07:45:02.006750Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning How To Ask: Cycle-Consistency Refines Prompts in Multimodal Foundation Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CV"],"primary_cat":"cs.CL","authors_text":"Gayathri Mahalingam, Jianzhe Lin, Jie Zhao, Maurice Diesendruck, Mingyang Xu, Shima Imani","submitted_at":"2024-02-13T19:49:17Z","abstract_excerpt":"When LLMs perform zero-shot inference, they typically use a prompt with a task specification, and generate a completion. However, there is no work to explore the possibility of the reverse - going from completion to task specification. In this paper, we employ both directions to perform cycle-supervised learning entirely in-context. Our goal is to create a forward map f : X -> Y (e.g. image -> generated caption), coupled with a backward map g : Y -> X (e.g. caption -> generated image) to construct a cycle-consistency \"loss\" (formulated as an update to the prompt) to enforce g(f(X)) ~= X. The t"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2402.08756","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/2402.08756/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":"2402.08756","created_at":"2026-07-05T07:45:02.006807+00:00"},{"alias_kind":"arxiv_version","alias_value":"2402.08756v1","created_at":"2026-07-05T07:45:02.006807+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2402.08756","created_at":"2026-07-05T07:45:02.006807+00:00"},{"alias_kind":"pith_short_12","alias_value":"N4VWJGE4AZSS","created_at":"2026-07-05T07:45:02.006807+00:00"},{"alias_kind":"pith_short_16","alias_value":"N4VWJGE4AZSS4BAL","created_at":"2026-07-05T07:45:02.006807+00:00"},{"alias_kind":"pith_short_8","alias_value":"N4VWJGE4","created_at":"2026-07-05T07:45:02.006807+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":1,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2505.24211","citing_title":"Are Any-to-Any Models More Consistent Across Modality Transfers Than Specialists?","ref_index":13,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ","json":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ.json","graph_json":"https://pith.science/api/pith-number/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/graph.json","events_json":"https://pith.science/api/pith-number/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/events.json","paper":"https://pith.science/paper/N4VWJGE4"},"agent_actions":{"view_html":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ","download_json":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ.json","view_paper":"https://pith.science/paper/N4VWJGE4","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2402.08756&json=true","fetch_graph":"https://pith.science/api/pith-number/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/graph.json","fetch_events":"https://pith.science/api/pith-number/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/action/timestamp_anchor","attest_storage":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/action/storage_attestation","attest_author":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/action/author_attestation","sign_citation":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/action/citation_signature","submit_replication":"https://pith.science/pith/N4VWJGE4AZSS4BAL2ZPV7A6CVZ/action/replication_record"}},"created_at":"2026-07-05T07:45:02.006807+00:00","updated_at":"2026-07-05T07:45:02.006807+00:00"}