{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:QMNI4LRTJL6LOOITEAK5OCLW7W","short_pith_number":"pith:QMNI4LRT","schema_version":"1.0","canonical_sha256":"831a8e2e334afcb739132015d70976fdadbe4be07b31c337a9fe286d353cebfd","source":{"kind":"arxiv","id":"2509.21552","version":2},"attestation_state":"computed","paper":{"title":"Learning GUI Grounding with Spatial Reasoning from Visual Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Chaoyun Zhang, Fangkai Yang, Huseyin Atahan Inan, Lukas Wutschitz, Lu Wang, Pasquale Minervini, Robert Sim, Samuel Kessler, Saravan Rajmohan, Wei-Ning Chen, Yu Zhao","submitted_at":"2025-09-25T20:38:01Z","abstract_excerpt":"Graphical User Interface (GUI) grounding is commonly framed as a coordinate prediction task -- given a natural language instruction, generate on-screen coordinates for actions such as clicks and keystrokes. However, recent Vision Language Models (VLMs) often fail to predict accurate numeric coordinates when processing GUI images with high resolutions and complex layouts. To address this issue, we reframe GUI grounding as an interactive search task, where the VLM generates actions to move a cursor in the GUI to locate UI elements. At each step, the model determines the target object, evaluates "},"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":"2509.21552","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.CV","submitted_at":"2025-09-25T20:38:01Z","cross_cats_sorted":["cs.CL"],"title_canon_sha256":"6e7428599cfad04afbd7a1d9e53e83b5b587b5ce7378657efe6a3b7dcab289f1","abstract_canon_sha256":"9b35883c64595c95709f0fbf0a978a390a3049142145fad5b78f8f81c95e03fc"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-27T01:04:51.056402Z","signature_b64":"j3jnRpcVKhU/dYbD9MBsOaxBohbZROjK1P++Yq0qUj0LMC7hOx4CRgeBBDU2ZJOJ6YVjvLxX5gz+hgddbRe1BQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"831a8e2e334afcb739132015d70976fdadbe4be07b31c337a9fe286d353cebfd","last_reissued_at":"2026-05-27T01:04:51.055642Z","signature_status":"signed_v1","first_computed_at":"2026-05-27T01:04:51.055642Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Learning GUI Grounding with Spatial Reasoning from Visual Feedback","license":"http://creativecommons.org/licenses/by/4.0/","headline":"","cross_cats":["cs.CL"],"primary_cat":"cs.CV","authors_text":"Chaoyun Zhang, Fangkai Yang, Huseyin Atahan Inan, Lukas Wutschitz, Lu Wang, Pasquale Minervini, Robert Sim, Samuel Kessler, Saravan Rajmohan, Wei-Ning Chen, Yu Zhao","submitted_at":"2025-09-25T20:38:01Z","abstract_excerpt":"Graphical User Interface (GUI) grounding is commonly framed as a coordinate prediction task -- given a natural language instruction, generate on-screen coordinates for actions such as clicks and keystrokes. However, recent Vision Language Models (VLMs) often fail to predict accurate numeric coordinates when processing GUI images with high resolutions and complex layouts. To address this issue, we reframe GUI grounding as an interactive search task, where the VLM generates actions to move a cursor in the GUI to locate UI elements. At each step, the model determines the target object, evaluates "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2509.21552","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/2509.21552/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":"2509.21552","created_at":"2026-05-27T01:04:51.055742+00:00"},{"alias_kind":"arxiv_version","alias_value":"2509.21552v2","created_at":"2026-05-27T01:04:51.055742+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2509.21552","created_at":"2026-05-27T01:04:51.055742+00:00"},{"alias_kind":"pith_short_12","alias_value":"QMNI4LRTJL6L","created_at":"2026-05-27T01:04:51.055742+00:00"},{"alias_kind":"pith_short_16","alias_value":"QMNI4LRTJL6LOOIT","created_at":"2026-05-27T01:04:51.055742+00:00"},{"alias_kind":"pith_short_8","alias_value":"QMNI4LRT","created_at":"2026-05-27T01:04:51.055742+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":5,"internal_anchor_count":5,"sample":[{"citing_arxiv_id":"2605.15542","citing_title":"DRS-GUI: Dynamic Region Search for Training-Free GUI Grounding","ref_index":53,"is_internal_anchor":true},{"citing_arxiv_id":"2605.12549","citing_title":"What Happens Before Decoding? Prefill Determines GUI Grounding in VLMs","ref_index":26,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25380","citing_title":"Benchmarking and Improving GUI Agents in High-Dynamic Environments","ref_index":50,"is_internal_anchor":true},{"citing_arxiv_id":"2604.21268","citing_title":"Measure Twice, Click Once: Co-evolving Proposer and Visual Critic via Reinforcement Learning for GUI Grounding","ref_index":68,"is_internal_anchor":true},{"citing_arxiv_id":"2604.25380","citing_title":"Benchmarking and Improving GUI Agents in High-Dynamic Environments","ref_index":50,"is_internal_anchor":true}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W","json":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W.json","graph_json":"https://pith.science/api/pith-number/QMNI4LRTJL6LOOITEAK5OCLW7W/graph.json","events_json":"https://pith.science/api/pith-number/QMNI4LRTJL6LOOITEAK5OCLW7W/events.json","paper":"https://pith.science/paper/QMNI4LRT"},"agent_actions":{"view_html":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W","download_json":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W.json","view_paper":"https://pith.science/paper/QMNI4LRT","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2509.21552&json=true","fetch_graph":"https://pith.science/api/pith-number/QMNI4LRTJL6LOOITEAK5OCLW7W/graph.json","fetch_events":"https://pith.science/api/pith-number/QMNI4LRTJL6LOOITEAK5OCLW7W/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W/action/timestamp_anchor","attest_storage":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W/action/storage_attestation","attest_author":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W/action/author_attestation","sign_citation":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W/action/citation_signature","submit_replication":"https://pith.science/pith/QMNI4LRTJL6LOOITEAK5OCLW7W/action/replication_record"}},"created_at":"2026-05-27T01:04:51.055742+00:00","updated_at":"2026-05-27T01:04:51.055742+00:00"}