The field of time series forecasting is moving towards incorporating more nuanced and dynamic approaches to modeling temporal dependencies. Recent research has highlighted the importance of learning latent hierarchical channel structures and exploiting cross-channel information to improve forecasting accuracy. Additionally, there is a growing recognition of the need for models that can learn the underlying dynamics of the data. Noteworthy papers include: ParallelTime, which proposes a dynamic weighting mechanism to balance short- and long-term dependencies, achieving state-of-the-art performance across diverse benchmarks. Are We Overlooking the Dimensions, which introduces U-Cast, a channel-dependent forecasting architecture that learns latent hierarchical channel structures, and releases Time-HD, a benchmark of large, diverse, high-dimensional datasets.