{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2024:XJ6HIL3O2Q2C6KWUT7LE4XDRXB","short_pith_number":"pith:XJ6HIL3O","schema_version":"1.0","canonical_sha256":"ba7c742f6ed4342f2ad49fd64e5c71b8409a28dc64d931cb839aebe14b28e7bf","source":{"kind":"arxiv","id":"2403.09193","version":2},"attestation_state":"computed","paper":{"title":"Can We Talk Models Into Seeing the World Differently?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","q-bio.NC"],"primary_cat":"cs.CV","authors_text":"Janis Keuper, Jovita Lukasik, Margret Keuper, M. Jehanzeb Mirza, Paul Gavrikov, Robert Geirhos, Steffen Jung","submitted_at":"2024-03-14T09:07:14Z","abstract_excerpt":"Unlike traditional vision-only models, vision language models (VLMs) offer an intuitive way to access visual content through language prompting by combining a large language model (LLM) with a vision encoder. However, both the LLM and the vision encoder come with their own set of biases, cue preferences, and shortcuts, which have been rigorously studied in uni-modal models. A timely question is how such (potentially misaligned) biases and cue preferences behave under multi-modal fusion in VLMs. As a first step towards a better understanding, we investigate a particularly well-studied vision-on"},"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":"2403.09193","kind":"arxiv","version":2},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.CV","submitted_at":"2024-03-14T09:07:14Z","cross_cats_sorted":["cs.AI","cs.LG","q-bio.NC"],"title_canon_sha256":"e4d029a3de4cdc18fc6852cb698417811da4cd5cc9866285706ceebaa6e13dfd","abstract_canon_sha256":"2444bbbf23457ae824d906fc64af5c111702b996c77d6bd8d10dddcf68c2c7cf"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T10:25:05.141882Z","signature_b64":"5BPidQnoL9meOhN29fugYNJWRU3UkqPw/E58ReM3hMvAaENSzANd8UP9Py1sHCBBxuFAe1yLRONXQOD0GRPKDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"ba7c742f6ed4342f2ad49fd64e5c71b8409a28dc64d931cb839aebe14b28e7bf","last_reissued_at":"2026-07-05T10:25:05.140930Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T10:25:05.140930Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Can We Talk Models Into Seeing the World Differently?","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.AI","cs.LG","q-bio.NC"],"primary_cat":"cs.CV","authors_text":"Janis Keuper, Jovita Lukasik, Margret Keuper, M. Jehanzeb Mirza, Paul Gavrikov, Robert Geirhos, Steffen Jung","submitted_at":"2024-03-14T09:07:14Z","abstract_excerpt":"Unlike traditional vision-only models, vision language models (VLMs) offer an intuitive way to access visual content through language prompting by combining a large language model (LLM) with a vision encoder. However, both the LLM and the vision encoder come with their own set of biases, cue preferences, and shortcuts, which have been rigorously studied in uni-modal models. A timely question is how such (potentially misaligned) biases and cue preferences behave under multi-modal fusion in VLMs. As a first step towards a better understanding, we investigate a particularly well-studied vision-on"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2403.09193","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/2403.09193/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":"2403.09193","created_at":"2026-07-05T10:25:05.141082+00:00"},{"alias_kind":"arxiv_version","alias_value":"2403.09193v2","created_at":"2026-07-05T10:25:05.141082+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2403.09193","created_at":"2026-07-05T10:25:05.141082+00:00"},{"alias_kind":"pith_short_12","alias_value":"XJ6HIL3O2Q2C","created_at":"2026-07-05T10:25:05.141082+00:00"},{"alias_kind":"pith_short_16","alias_value":"XJ6HIL3O2Q2C6KWU","created_at":"2026-07-05T10:25:05.141082+00:00"},{"alias_kind":"pith_short_8","alias_value":"XJ6HIL3O","created_at":"2026-07-05T10:25:05.141082+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":2,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2506.09082","citing_title":"AVA-Bench: Atomic Visual Ability Benchmark for Vision Foundation Models","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2604.10999","citing_title":"TraversalBench: Challenging Paths to Follow for Vision Language Models","ref_index":12,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB","json":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB.json","graph_json":"https://pith.science/api/pith-number/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/graph.json","events_json":"https://pith.science/api/pith-number/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/events.json","paper":"https://pith.science/paper/XJ6HIL3O"},"agent_actions":{"view_html":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB","download_json":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB.json","view_paper":"https://pith.science/paper/XJ6HIL3O","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2403.09193&json=true","fetch_graph":"https://pith.science/api/pith-number/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/graph.json","fetch_events":"https://pith.science/api/pith-number/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/action/timestamp_anchor","attest_storage":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/action/storage_attestation","attest_author":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/action/author_attestation","sign_citation":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/action/citation_signature","submit_replication":"https://pith.science/pith/XJ6HIL3O2Q2C6KWUT7LE4XDRXB/action/replication_record"}},"created_at":"2026-07-05T10:25:05.141082+00:00","updated_at":"2026-07-05T10:25:05.141082+00:00"}