{"paper":{"title":"CodeGraphVLP: Code-as-Planner Meets Semantic-Graph State for Non-Markovian Vision-Language-Action Models","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"CodeGraphVLP pairs a persistent semantic-graph state with a code planner to succeed more often on non-Markovian robot tasks than standard VLA models.","cross_cats":[],"primary_cat":"cs.RO","authors_text":"Anh Nguyen, Anthony Gunderman, Chase Rainwater, Duy Nguyen, Khoa Vo, Minh Vu, Ngan Le, Nghi D. Q. Bui, Sieu Tran, Taisei Hanyu, Yuki Ikebe","submitted_at":"2026-04-24T05:27:27Z","abstract_excerpt":"Vision-Language-Action (VLA) models promise generalist robot manipulation, but are typically trained and deployed as short-horizon policies that assume the latest observation is sufficient for action reasoning. This assumption breaks in non-Markovian long-horizon tasks, where task-relevant evidence can be occluded or appear only earlier in the trajectory, and where clutter and distractors make fine-grained visual grounding brittle. We present CodeGraphVLP, a hierarchical framework that enables reliable long-horizon manipulation by combining a persistent semantic-graph state with an executable "},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"On real-world non-Markovian tasks, CodeGraphVLP improves task completion over strong VLA baselines and history-enabled variants while substantially lowering planning latency compared to VLM-in-the-loop planning.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The semantic-graph state can reliably maintain task-relevant entities and relations under partial observability in cluttered real environments, allowing the code planner to perform accurate progress checks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"CodeGraphVLP uses a semantic-graph state and executable code planner to enable reliable long-horizon non-Markovian robot manipulation, improving task success and lowering latency over standard VLA baselines.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"CodeGraphVLP pairs a persistent semantic-graph state with a code planner to succeed more often on non-Markovian robot tasks than standard VLA models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"09fe828855e45a9d9c2c161890c47742bc1522eb56674ccc8fab7a63e9b655f5"},"source":{"id":"2604.22238","kind":"arxiv","version":2},"verdict":{"id":"69089529-7357-4ced-807a-a6354cefc4d8","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T11:37:49.770256Z","strongest_claim":"On real-world non-Markovian tasks, CodeGraphVLP improves task completion over strong VLA baselines and history-enabled variants while substantially lowering planning latency compared to VLM-in-the-loop planning.","one_line_summary":"CodeGraphVLP uses a semantic-graph state and executable code planner to enable reliable long-horizon non-Markovian robot manipulation, improving task success and lowering latency over standard VLA baselines.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The semantic-graph state can reliably maintain task-relevant entities and relations under partial observability in cluttered real environments, allowing the code planner to perform accurate progress checks.","pith_extraction_headline":"CodeGraphVLP pairs a persistent semantic-graph state with a code planner to succeed more often on non-Markovian robot tasks than standard VLA models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.22238/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T11:34:28.423744Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T00:12:04.628799Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"85b2985d552e06cbbf0c1dcd1c907c033e7a520a48addd48c89d593eef1fcd1d"},"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"}