{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:BAWL5HRBNPUQU2N3NVILTQIMLU","short_pith_number":"pith:BAWL5HRB","schema_version":"1.0","canonical_sha256":"082cbe9e216be90a69bb6d50b9c10c5d12c1084c0203174af6b70c8298a1c3eb","source":{"kind":"arxiv","id":"2607.20351","version":1},"attestation_state":"computed","paper":{"title":"Test-Time Training for Modality Order Consistency in Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Aditi Gupta, Yossi Gandelsman","submitted_at":"2026-07-22T16:37:02Z","abstract_excerpt":"We find that vision-language models are sensitive to a specific semantically irrelevant change: the order in which the image and question are presented. Across three models and three benchmarks, image first prompting consistently outperforms question-first prompting, revealing a repeatable modality order failure. We use this gap to design an order-consistent test-time training method. Our method substantially closes the modality-order gap across all evaluated settings. Surprisingly, it also yields consistent improvements in the stronger image-first branch over the baseline, hence bootstrapping"},"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":"2607.20351","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-07-22T16:37:02Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"820319a6b7edc3325ef54b6d359544db9e49fcc16db0ccc1161731076fdb9704","abstract_canon_sha256":"e295d3e9953307792926d2968930c3a7e343604ec17c068d7da4f2304fb48f03"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-23T01:25:15.843964Z","signature_b64":"30g3MkuGx43C86uJjx+sCKpyWhmDrORFs66WVmpCRen88ADxrwZb9SfWfExx3KiUlf8XzkJRdwEZzlDMOdLeCA==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"082cbe9e216be90a69bb6d50b9c10c5d12c1084c0203174af6b70c8298a1c3eb","last_reissued_at":"2026-07-23T01:25:15.843078Z","signature_status":"signed_v1","first_computed_at":"2026-07-23T01:25:15.843078Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Test-Time Training for Modality Order Consistency in Vision-Language Models","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Aditi Gupta, Yossi Gandelsman","submitted_at":"2026-07-22T16:37:02Z","abstract_excerpt":"We find that vision-language models are sensitive to a specific semantically irrelevant change: the order in which the image and question are presented. Across three models and three benchmarks, image first prompting consistently outperforms question-first prompting, revealing a repeatable modality order failure. We use this gap to design an order-consistent test-time training method. Our method substantially closes the modality-order gap across all evaluated settings. Surprisingly, it also yields consistent improvements in the stronger image-first branch over the baseline, hence bootstrapping"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2607.20351","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/2607.20351/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":"2607.20351","created_at":"2026-07-23T01:25:15.843510+00:00"},{"alias_kind":"arxiv_version","alias_value":"2607.20351v1","created_at":"2026-07-23T01:25:15.843510+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2607.20351","created_at":"2026-07-23T01:25:15.843510+00:00"},{"alias_kind":"pith_short_12","alias_value":"BAWL5HRBNPUQ","created_at":"2026-07-23T01:25:15.843510+00:00"},{"alias_kind":"pith_short_16","alias_value":"BAWL5HRBNPUQU2N3","created_at":"2026-07-23T01:25:15.843510+00:00"},{"alias_kind":"pith_short_8","alias_value":"BAWL5HRB","created_at":"2026-07-23T01:25:15.843510+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":0,"internal_anchor_count":0,"sample":[]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU","json":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU.json","graph_json":"https://pith.science/api/pith-number/BAWL5HRBNPUQU2N3NVILTQIMLU/graph.json","events_json":"https://pith.science/api/pith-number/BAWL5HRBNPUQU2N3NVILTQIMLU/events.json","paper":"https://pith.science/paper/BAWL5HRB"},"agent_actions":{"view_html":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU","download_json":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU.json","view_paper":"https://pith.science/paper/BAWL5HRB","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2607.20351&json=true","fetch_graph":"https://pith.science/api/pith-number/BAWL5HRBNPUQU2N3NVILTQIMLU/graph.json","fetch_events":"https://pith.science/api/pith-number/BAWL5HRBNPUQU2N3NVILTQIMLU/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU/action/timestamp_anchor","attest_storage":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU/action/storage_attestation","attest_author":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU/action/author_attestation","sign_citation":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU/action/citation_signature","submit_replication":"https://pith.science/pith/BAWL5HRBNPUQU2N3NVILTQIMLU/action/replication_record"}},"created_at":"2026-07-23T01:25:15.843510+00:00","updated_at":"2026-07-23T01:25:15.843510+00:00"}