{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2022:JJDJILSUDCCKFD2XRYRVNUSYF4","short_pith_number":"pith:JJDJILSU","schema_version":"1.0","canonical_sha256":"4a46942e541884a28f578e2356d2582f0d9de957d721f2aa10332cd8b78b220e","source":{"kind":"arxiv","id":"2209.15162","version":3},"attestation_state":"computed","paper":{"title":"Linearly Mapping from Image to Text Space","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Carsten Eickhoff, Ellie Pavlick, Jack Merullo, Louis Castricato","submitted_at":"2022-09-30T01:17:18Z","abstract_excerpt":"The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by traini"},"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":"2209.15162","kind":"arxiv","version":3},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CL","submitted_at":"2022-09-30T01:17:18Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"4890d8cbee3de6dc4a26188b6ed46cb057eff06554adef18d8ad57ffda509b64","abstract_canon_sha256":"fe159d20d94cbb8f7cf33740e9c6f09846c00dde73a78f4b0fa3551d18e45b84"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T05:49:30.291633Z","signature_b64":"NZoPIK8/4b0p262ufxGougiwHPdI0d6VukclV/xgTrwZgBzkx7ghvKUjOmqJo0s8YtE4miFHu6ApapeYwnHYDA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"4a46942e541884a28f578e2356d2582f0d9de957d721f2aa10332cd8b78b220e","last_reissued_at":"2026-07-05T05:49:30.291183Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T05:49:30.291183Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Linearly Mapping from Image to Text Space","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.CL","authors_text":"Carsten Eickhoff, Ellie Pavlick, Jack Merullo, Louis Castricato","submitted_at":"2022-09-30T01:17:18Z","abstract_excerpt":"The extent to which text-only language models (LMs) learn to represent features of the non-linguistic world is an open question. Prior work has shown that pretrained LMs can be taught to caption images when a vision model's parameters are optimized to encode images in the language space. We test a stronger hypothesis: that the conceptual representations learned by frozen text-only models and vision-only models are similar enough that this can be achieved with a linear map. We show that the image representations from vision models can be transferred as continuous prompts to frozen LMs by traini"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2209.15162","kind":"arxiv","version":3},"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/2209.15162/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":"2209.15162","created_at":"2026-07-05T05:49:30.291241+00:00"},{"alias_kind":"arxiv_version","alias_value":"2209.15162v3","created_at":"2026-07-05T05:49:30.291241+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2209.15162","created_at":"2026-07-05T05:49:30.291241+00:00"},{"alias_kind":"pith_short_12","alias_value":"JJDJILSUDCCK","created_at":"2026-07-05T05:49:30.291241+00:00"},{"alias_kind":"pith_short_16","alias_value":"JJDJILSUDCCKFD2X","created_at":"2026-07-05T05:49:30.291241+00:00"},{"alias_kind":"pith_short_8","alias_value":"JJDJILSU","created_at":"2026-07-05T05:49:30.291241+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":7,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.20077","citing_title":"The Hidden Evolution of Disguised Visual Context inside the VLM","ref_index":56,"is_internal_anchor":false},{"citing_arxiv_id":"2606.06021","citing_title":"OPRD: On-Policy Representation Distillation","ref_index":26,"is_internal_anchor":false},{"citing_arxiv_id":"2605.16834","citing_title":"Learning Relative Representations for Fine-Grained Multimodal Alignment with Limited Data","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2405.07987","citing_title":"The Platonic Representation Hypothesis","ref_index":284,"is_internal_anchor":false},{"citing_arxiv_id":"2604.02486","citing_title":"VLMs Need Words: Vision Language Models Ignore Visual Detail In Favor of Semantic Anchors","ref_index":10,"is_internal_anchor":false},{"citing_arxiv_id":"2605.12412","citing_title":"Stories in Space: In-Context Learning Trajectories in Conceptual Belief Space","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2605.06640","citing_title":"Concept-Based Abductive and Contrastive Explanations for Behaviors of Vision Models","ref_index":74,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4","json":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4.json","graph_json":"https://pith.science/api/pith-number/JJDJILSUDCCKFD2XRYRVNUSYF4/graph.json","events_json":"https://pith.science/api/pith-number/JJDJILSUDCCKFD2XRYRVNUSYF4/events.json","paper":"https://pith.science/paper/JJDJILSU"},"agent_actions":{"view_html":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4","download_json":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4.json","view_paper":"https://pith.science/paper/JJDJILSU","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2209.15162&json=true","fetch_graph":"https://pith.science/api/pith-number/JJDJILSUDCCKFD2XRYRVNUSYF4/graph.json","fetch_events":"https://pith.science/api/pith-number/JJDJILSUDCCKFD2XRYRVNUSYF4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4/action/storage_attestation","attest_author":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4/action/author_attestation","sign_citation":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4/action/citation_signature","submit_replication":"https://pith.science/pith/JJDJILSUDCCKFD2XRYRVNUSYF4/action/replication_record"}},"created_at":"2026-07-05T05:49:30.291241+00:00","updated_at":"2026-07-05T05:49:30.291241+00:00"}