{"paper":{"title":"SpaMEM: Benchmarking Dynamic Spatial Reasoning via Perception-Memory Integration in Embodied Environments","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"Vision-language models maintain spatial beliefs only when given text histories and collapse without them in changing environments.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Chih-Ting Liao, Chunlei Meng, Tianyang Wang, Weilin Zhou, Xin Cao, Xi Xiao, Xu Zheng, Yitong Qiao, Zhangquan Chen","submitted_at":"2026-04-24T10:06:41Z","abstract_excerpt":"Multimodal large language models (MLLMs) have advanced static visual--spatial reasoning, yet they often fail to preserve long-horizon spatial coherence in embodied settings where beliefs must be continuously revised from egocentric observations under environmental change. We introduce SpaMEM (Spatial Memory from Action Sequences), a large-scale diagnostic benchmark that isolates the mechanics of spatial belief evolution via action-conditioned scene transformations (spawn, place, remove) over long interaction horizons. SpaMEM is built on a physically grounded dataset with 10,601,392 high-fideli"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"Benchmarking representative open-source VLM families reveals a consistent stacked bottleneck: coordinate-consistent grounding remains a hard ceiling, and the sharp collapse from Level 2 to Level 3 exposes a pronounced symbolic scaffolding dependency, where models succeed with text-based bookkeeping but struggle to sustain robust visual memory.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the procedurally generated houses, action sequences (spawn, place, remove), and three-level task hierarchy isolate spatial belief evolution without introducing simulation artifacts or task-specific biases that do not appear in real embodied settings.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"SpaMEM benchmark shows multimodal LLMs succeed at spatial tasks with text histories but sharply fail at long-horizon belief maintenance from raw visual streams alone.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"Vision-language models maintain spatial beliefs only when given text histories and collapse without them in changing environments.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"8097cc1a1a6efda0760862e192342df6de894f45cbfe100507443f4f6d405811"},"source":{"id":"2604.22409","kind":"arxiv","version":2},"verdict":{"id":"cdf74359-034b-4e8a-bb2b-161706349604","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-08T12:29:34.022871Z","strongest_claim":"Benchmarking representative open-source VLM families reveals a consistent stacked bottleneck: coordinate-consistent grounding remains a hard ceiling, and the sharp collapse from Level 2 to Level 3 exposes a pronounced symbolic scaffolding dependency, where models succeed with text-based bookkeeping but struggle to sustain robust visual memory.","one_line_summary":"SpaMEM benchmark shows multimodal LLMs succeed at spatial tasks with text histories but sharply fail at long-horizon belief maintenance from raw visual streams alone.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the procedurally generated houses, action sequences (spawn, place, remove), and three-level task hierarchy isolate spatial belief evolution without introducing simulation artifacts or task-specific biases that do not appear in real embodied settings.","pith_extraction_headline":"Vision-language models maintain spatial beliefs only when given text histories and collapse without them in changing environments."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2604.22409/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"ai_meta_artifact","ran_at":"2026-05-21T10:40:36.394346Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-20T00:03:40.140806Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"67bf1259c6a25510706a17a3d92f5f87875235c59a524d485cd6e2c8addbc6e4"},"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"}