Tag
LongCrafter proposes a structured synthesis framework that uses a hierarchical task taxonomy and evidence graphs to generate diverse, faithful long-context instruction data, significantly improving LLM long-context understanding.
Introduces Φ-Nav, a unified on-policy framework that uses hindsight reasoning to synthetically generate path-level instructions from exploratory trajectories, bridging the semantic supervision gap in Vision-Language Navigation and achieving competitive results on R2R-CE and RxR-CE benchmarks with fewer expert demonstrations.