{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2023:GVUBWLVXBT5BTL7VMXFPJZQL2Y","short_pith_number":"pith:GVUBWLVX","schema_version":"1.0","canonical_sha256":"35681b2eb70cfa19aff565caf4e60bd622b335be6081eeec8a43d86496a1f2ea","source":{"kind":"arxiv","id":"2306.01879","version":4},"attestation_state":"computed","paper":{"title":"Revisiting the Role of Language Priors in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Deepak Pathak, Deva Ramanan, Pengchuan Zhang, Xinyue Chen, Zhiqiu Lin","submitted_at":"2023-06-02T19:19:43Z","abstract_excerpt":"Vision-language models (VLMs) are impactful in part because they can be applied to a variety of visual understanding tasks in a zero-shot fashion, without any fine-tuning. We study $\\textit{generative VLMs}$ that are trained for next-word generation given an image. We explore their zero-shot performance on the illustrative task of image-text retrieval across 8 popular vision-language benchmarks. Our first observation is that they can be repurposed for discriminative tasks (such as image-text retrieval) by simply computing the match score of generating a particular text string given an image. W"},"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":"2306.01879","kind":"arxiv","version":4},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2023-06-02T19:19:43Z","cross_cats_sorted":["cs.AI","cs.CL"],"title_canon_sha256":"b8c42682a64873acc7fafdbe88829a9e4e5bd0a6e5ece7c5c92b586e28003264","abstract_canon_sha256":"03b96a90594c99c1c2282d028437027f640cca20f944b2413933eb864eecda42"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T08:19:19.435174Z","signature_b64":"hHIC59X/CvAKf+paTMHzbjd2MRFjqkJuG56R3lC8uZjtrjgaT3DIEOZiF+J++4nuu3bouKTCRrUei+JyOqHCBQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"35681b2eb70cfa19aff565caf4e60bd622b335be6081eeec8a43d86496a1f2ea","last_reissued_at":"2026-07-05T08:19:19.434679Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T08:19:19.434679Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Revisiting the Role of Language Priors in Vision-Language Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.CL"],"primary_cat":"cs.CV","authors_text":"Deepak Pathak, Deva Ramanan, Pengchuan Zhang, Xinyue Chen, Zhiqiu Lin","submitted_at":"2023-06-02T19:19:43Z","abstract_excerpt":"Vision-language models (VLMs) are impactful in part because they can be applied to a variety of visual understanding tasks in a zero-shot fashion, without any fine-tuning. We study $\\textit{generative VLMs}$ that are trained for next-word generation given an image. We explore their zero-shot performance on the illustrative task of image-text retrieval across 8 popular vision-language benchmarks. Our first observation is that they can be repurposed for discriminative tasks (such as image-text retrieval) by simply computing the match score of generating a particular text string given an image. W"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2306.01879","kind":"arxiv","version":4},"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/2306.01879/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":"2306.01879","created_at":"2026-07-05T08:19:19.434742+00:00"},{"alias_kind":"arxiv_version","alias_value":"2306.01879v4","created_at":"2026-07-05T08:19:19.434742+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2306.01879","created_at":"2026-07-05T08:19:19.434742+00:00"},{"alias_kind":"pith_short_12","alias_value":"GVUBWLVXBT5B","created_at":"2026-07-05T08:19:19.434742+00:00"},{"alias_kind":"pith_short_16","alias_value":"GVUBWLVXBT5BTL7V","created_at":"2026-07-05T08:19:19.434742+00:00"},{"alias_kind":"pith_short_8","alias_value":"GVUBWLVX","created_at":"2026-07-05T08:19:19.434742+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2605.21919","citing_title":"SDGBiasBench: Benchmarking and Mitigating Vision--Language Models' Biases in Sustainable Development Goals","ref_index":34,"is_internal_anchor":false},{"citing_arxiv_id":"2604.21718","citing_title":"Building a Precise Video Language with Human-AI Oversight","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y","json":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y.json","graph_json":"https://pith.science/api/pith-number/GVUBWLVXBT5BTL7VMXFPJZQL2Y/graph.json","events_json":"https://pith.science/api/pith-number/GVUBWLVXBT5BTL7VMXFPJZQL2Y/events.json","paper":"https://pith.science/paper/GVUBWLVX"},"agent_actions":{"view_html":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y","download_json":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y.json","view_paper":"https://pith.science/paper/GVUBWLVX","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2306.01879&json=true","fetch_graph":"https://pith.science/api/pith-number/GVUBWLVXBT5BTL7VMXFPJZQL2Y/graph.json","fetch_events":"https://pith.science/api/pith-number/GVUBWLVXBT5BTL7VMXFPJZQL2Y/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y/action/timestamp_anchor","attest_storage":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y/action/storage_attestation","attest_author":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y/action/author_attestation","sign_citation":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y/action/citation_signature","submit_replication":"https://pith.science/pith/GVUBWLVXBT5BTL7VMXFPJZQL2Y/action/replication_record"}},"created_at":"2026-07-05T08:19:19.434742+00:00","updated_at":"2026-07-05T08:19:19.434742+00:00"}