[논문리뷰] GUI-Primitives: Diagnosing Spatial Reasoning Failures in Vision-Language GUI GroundingComputer-use agents는 screenshot을 기반으로 natural-language instructions를 GUI에 grounding하여 interface elements를 찾아냅니다.#Review#GUI Grounding#Spatial Reasoning#Vision-Language Models#Minimal-Pair Evaluation#Computer-Use Agents#Diagnostic Benchmark#Inference-Time Interventions2026년 8월 26일댓글 수 로딩 중