Search papers, labs, and topics across Lattice.
This paper introduces Q-GeoMem, a question-guided geometric memory framework for video spatial reasoning that addresses the limitations of treating memory as a generic temporal cache. Q-GeoMem maintains two complementary memories: a Fine-Grained Context Bank for recent dense features and camera states, and a Semantic-Geometric Evidence Bank for compact long-range evidence, updated based on question relevance and novelty. Experiments on VSI-Bench and VSTI-Bench demonstrate state-of-the-art performance, validating the effectiveness of question-guided geometric memory for improved long-horizon reasoning.
Forget generic video memory caches – Q-GeoMem selectively remembers only the geometrically relevant bits, boosting spatial reasoning performance.
Video spatial reasoning requires accumulating viewpoint-dependent evidence over time while retaining information useful to the question being asked. Existing spatial video-language models improve geometric perception and long-range context modeling, but often treat memory as a generic temporal cache, which can introduce redundant or irrelevant geometry and weaken long-horizon reasoning. We propose \textbf{\ours}, a question-guided geometric memory framework for video spatial reasoning. \ours injects camera-conditioned geometry into visual tokens and maintains two complementary memories: a Fine-Grained Context Bank for recent dense features and camera states, and a Semantic-Geometric Evidence Bank for compact long-range evidence. Each candidate frame is scored by the product of Q-Former-based question relevance and novelty with respect to the retained bank; this score is stored and reused during reading, while a capacity-based replacement rule keeps the bank compact. During reasoning, both memories are read before update and adaptively fused with the current frame representation. Experiments on VSI-Bench and VSTI-Bench show that \ours achieves state-of-the-art performance among evaluated spatial reasoning models, validating the effectiveness of question-guided geometric memory. Ablations further verify the contribution of the proposed evidence scoring mechanism.
