{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2026:6AVSL3D6IDIFXNW3WIRUZTW6E4","short_pith_number":"pith:6AVSL3D6","schema_version":"1.0","canonical_sha256":"f02b25ec7e40d05bb6dbb2234ccede272e5562e60dcf2efd01f819370cac98d1","source":{"kind":"arxiv","id":"2606.13156","version":1},"attestation_state":"computed","paper":{"title":"Iterative Visual Thinking: Teaching Vision-Language Models Spatial Self-Correction through Visual Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Animesh Tripathy, Aswanth Krishnan","submitted_at":"2026-06-11T10:27:29Z","abstract_excerpt":"Vision-language models (VLMs) achieve strong singleshot spatial grounding, yet lack any mechanism to observe and correct their own predictions. We find that naively prompting a VLM to iterate over rendered visualizations of its predictions causes catastrophic failure: Acc@0.5 on referring expression comprehension collapses from 79.6% to 48.7% (a 31 percentage point drop), revealing a fundamental gap between grounding capability and self-correction ability. We propose Iterative Visual Thinking (IVT), a closed-loop framework in which the model predicts a bounding box, observes the prediction ren"},"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":"2606.13156","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2026-06-11T10:27:29Z","cross_cats_sorted":["cs.AI"],"title_canon_sha256":"9522d90b78b706860a9ff0e000ce52fa8773c1c863c7b526f50a01d7d8c3320f","abstract_canon_sha256":"967bec805f6e53ea2a7751ffc111102b3a1244a7b460cb6d640b71300629381c"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-06-12T01:09:43.895147Z","signature_b64":"ejUiKcYKeQCZFVBrlPiHeUe8fEMrSVJ0i1VAmpXr4UpiY8BpMMoA7Z+4u62w0I0j7irWQEKLDIx2iGbrXsyXDw==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"f02b25ec7e40d05bb6dbb2234ccede272e5562e60dcf2efd01f819370cac98d1","last_reissued_at":"2026-06-12T01:09:43.894178Z","signature_status":"signed_v1","first_computed_at":"2026-06-12T01:09:43.894178Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Iterative Visual Thinking: Teaching Vision-Language Models Spatial Self-Correction through Visual Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.AI"],"primary_cat":"cs.CV","authors_text":"Animesh Tripathy, Aswanth Krishnan","submitted_at":"2026-06-11T10:27:29Z","abstract_excerpt":"Vision-language models (VLMs) achieve strong singleshot spatial grounding, yet lack any mechanism to observe and correct their own predictions. We find that naively prompting a VLM to iterate over rendered visualizations of its predictions causes catastrophic failure: Acc@0.5 on referring expression comprehension collapses from 79.6% to 48.7% (a 31 percentage point drop), revealing a fundamental gap between grounding capability and self-correction ability. We propose Iterative Visual Thinking (IVT), a closed-loop framework in which the model predicts a bounding box, observes the prediction ren"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2606.13156","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/2606.13156/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":"2606.13156","created_at":"2026-06-12T01:09:43.894297+00:00"},{"alias_kind":"arxiv_version","alias_value":"2606.13156v1","created_at":"2026-06-12T01:09:43.894297+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2606.13156","created_at":"2026-06-12T01:09:43.894297+00:00"},{"alias_kind":"pith_short_12","alias_value":"6AVSL3D6IDIF","created_at":"2026-06-12T01:09:43.894297+00:00"},{"alias_kind":"pith_short_16","alias_value":"6AVSL3D6IDIFXNW3","created_at":"2026-06-12T01:09:43.894297+00:00"},{"alias_kind":"pith_short_8","alias_value":"6AVSL3D6","created_at":"2026-06-12T01:09:43.894297+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/6AVSL3D6IDIFXNW3WIRUZTW6E4","json":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4.json","graph_json":"https://pith.science/api/pith-number/6AVSL3D6IDIFXNW3WIRUZTW6E4/graph.json","events_json":"https://pith.science/api/pith-number/6AVSL3D6IDIFXNW3WIRUZTW6E4/events.json","paper":"https://pith.science/paper/6AVSL3D6"},"agent_actions":{"view_html":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4","download_json":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4.json","view_paper":"https://pith.science/paper/6AVSL3D6","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2606.13156&json=true","fetch_graph":"https://pith.science/api/pith-number/6AVSL3D6IDIFXNW3WIRUZTW6E4/graph.json","fetch_events":"https://pith.science/api/pith-number/6AVSL3D6IDIFXNW3WIRUZTW6E4/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4/action/timestamp_anchor","attest_storage":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4/action/storage_attestation","attest_author":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4/action/author_attestation","sign_citation":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4/action/citation_signature","submit_replication":"https://pith.science/pith/6AVSL3D6IDIFXNW3WIRUZTW6E4/action/replication_record"}},"created_at":"2026-06-12T01:09:43.894297+00:00","updated_at":"2026-06-12T01:09:43.894297+00:00"}