{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2021:ELCU4XF562WNZJ5OJHUIHJOPJH","short_pith_number":"pith:ELCU4XF5","schema_version":"1.0","canonical_sha256":"22c54e5cbdf6acdca7ae49e883a5cf49f07ca56e3010017736ce4371d004ab01","source":{"kind":"arxiv","id":"2106.13884","version":2},"attestation_state":"computed","paper":{"title":"Multimodal Few-Shot Learning with Frozen Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Felix Hill, Jacob Menick, Maria Tsimpoukelli, Oriol Vinyals, Serkan Cabi, S. M. Ali Eslami","submitted_at":"2021-06-25T21:07:09Z","abstract_excerpt":"When trained at sufficient scale, auto-regressive language models exhibit the notable ability to learn a new language task after being prompted with just a few examples. Here, we present a simple, yet effective, approach for transferring this few-shot learning ability to a multimodal setting (vision and language). Using aligned image and caption data, we train a vision encoder to represent each image as a sequence of continuous embeddings, such that a pre-trained, frozen language model prompted with this prefix generates the appropriate caption. The resulting system is a multimodal few-shot le"},"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":"2106.13884","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2021-06-25T21:07:09Z","cross_cats_sorted":["cs.CL","cs.LG"],"title_canon_sha256":"953a5ecb14c97fc1eac42807e1d98f66d670eaec7744bf3f41fbf0506e6afdd4","abstract_canon_sha256":"44806c2f7fbed3bb9d542fa4dcc8ab03627f10e22035186d0d5aba065ea29f73"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T02:54:40.593532Z","signature_b64":"WT2HpG0AYCPZdbqG5RuzWjDIe3CgTnVgk/9ksmpQLrL5lQ5t20BPP8zky/7BT3vjwUIZ02YEH+IRyIc/aSWJAA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"22c54e5cbdf6acdca7ae49e883a5cf49f07ca56e3010017736ce4371d004ab01","last_reissued_at":"2026-07-05T02:54:40.593048Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T02:54:40.593048Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Multimodal Few-Shot Learning with Frozen Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.CL","cs.LG"],"primary_cat":"cs.CV","authors_text":"Felix Hill, Jacob Menick, Maria Tsimpoukelli, Oriol Vinyals, Serkan Cabi, S. M. Ali Eslami","submitted_at":"2021-06-25T21:07:09Z","abstract_excerpt":"When trained at sufficient scale, auto-regressive language models exhibit the notable ability to learn a new language task after being prompted with just a few examples. Here, we present a simple, yet effective, approach for transferring this few-shot learning ability to a multimodal setting (vision and language). Using aligned image and caption data, we train a vision encoder to represent each image as a sequence of continuous embeddings, such that a pre-trained, frozen language model prompted with this prefix generates the appropriate caption. The resulting system is a multimodal few-shot le"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2106.13884","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/2106.13884/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":"2106.13884","created_at":"2026-07-05T02:54:40.593102+00:00"},{"alias_kind":"arxiv_version","alias_value":"2106.13884v2","created_at":"2026-07-05T02:54:40.593102+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2106.13884","created_at":"2026-07-05T02:54:40.593102+00:00"},{"alias_kind":"pith_short_12","alias_value":"ELCU4XF562WN","created_at":"2026-07-05T02:54:40.593102+00:00"},{"alias_kind":"pith_short_16","alias_value":"ELCU4XF562WNZJ5O","created_at":"2026-07-05T02:54:40.593102+00:00"},{"alias_kind":"pith_short_8","alias_value":"ELCU4XF5","created_at":"2026-07-05T02:54:40.593102+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":1,"sample":[{"citing_arxiv_id":"2607.06445","citing_title":"Analysis-by-Proxy: Localization Signals in VLMs Operating as Condition Encoders","ref_index":24,"is_internal_anchor":true},{"citing_arxiv_id":"2303.11381","citing_title":"MM-REACT: Prompting ChatGPT for Multimodal Reasoning and Action","ref_index":27,"is_internal_anchor":false},{"citing_arxiv_id":"2407.07726","citing_title":"PaliGemma: A versatile 3B VLM for transfer","ref_index":134,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH","json":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH.json","graph_json":"https://pith.science/api/pith-number/ELCU4XF562WNZJ5OJHUIHJOPJH/graph.json","events_json":"https://pith.science/api/pith-number/ELCU4XF562WNZJ5OJHUIHJOPJH/events.json","paper":"https://pith.science/paper/ELCU4XF5"},"agent_actions":{"view_html":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH","download_json":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH.json","view_paper":"https://pith.science/paper/ELCU4XF5","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2106.13884&json=true","fetch_graph":"https://pith.science/api/pith-number/ELCU4XF562WNZJ5OJHUIHJOPJH/graph.json","fetch_events":"https://pith.science/api/pith-number/ELCU4XF562WNZJ5OJHUIHJOPJH/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH/action/timestamp_anchor","attest_storage":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH/action/storage_attestation","attest_author":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH/action/author_attestation","sign_citation":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH/action/citation_signature","submit_replication":"https://pith.science/pith/ELCU4XF562WNZJ5OJHUIHJOPJH/action/replication_record"}},"created_at":"2026-07-05T02:54:40.593102+00:00","updated_at":"2026-07-05T02:54:40.593102+00:00"}