When Does Latent Reasoning Help? MeRa: Metric-Space Bias for Spatial Prediction
For researchers in spatial prediction and sequential recommendation, this work identifies a critical condition (metric-space grounding) for effective latent reasoning, offering a simple plug-in module that improves performance.
Latent reasoning degrades spatial prediction without grounding in the underlying metric space, but adding a learned metric-space bias (MeRa) yields consistent gains, achieving up to 4.5% NDCG@10 improvement over unmodified baselines and surpassing recent methods like GeoMamba and HMST on all three benchmarks.
Latent reasoning has improved sequential recommendation by iteratively refining representations before prediction, but does it help spatial prediction? We find that the answer depends on whether reasoning is grounded in the underlying metric space. Without such grounding, latent reasoning degrades spatial prediction below the unmodified baseline, while a learned metric-space bias derived from pairwise distances produces consistent gains. We formalize this finding through MeRa (Metric-space Reasoning), a lightweight backbone-agnostic module that can be inserted between any sequence encoder and its prediction heads. On the GETNext backbone, the gap between reasoning without and with metric-space bias reaches 4.5% NDCG@10. MeRa achieves the best NDCG@10 on all three spatial prediction benchmarks among the compared methods, surpassing recent approaches such as GeoMamba and HMST. We prove that metric-space-constrained reasoning converges to a unique fixed point and that N-step reasoning is strictly more expressive than (N-1)-step reasoning. A controlled experiment on CLEVR with Euclidean distance confirms that the finding generalizes beyond geographic coordinates. The code is included in the supplementary material.