AI RESEARCH
StateLinFormer: Stateful Training Enhancing Long-term Memory in Navigation
arXiv CS.LG
•
ArXi:2603.23571v1 Announce Type: new Effective navigation intelligence relies on long-term memory to both immediate generalization and sustained adaptation. However, existing approaches face a dilemma: modular systems rely on explicit mapping but lack flexibility, while Transformer-based end-to-end models are constrained by fixed context windows, limiting persistent memory across extended interactions. We