Reproduce “NeST: Neighborhood-aware semantic alignment and temporal modulation for LLM based time series forecasting” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jayanie Bogahawatte, Sachith Seneviratne, Maneesha Perera, Saman Halgamuge
NeST uses learned text prototypes to modulate time-series features rather than adding semantic tokens to an LLM's input. It learns prototypes in pretrained word-embedding space, aligns them with time-series representations through nearest-neighbor contrastive learning, and uses retrieved prototypes to scale and shift patches. On ETTh1, its average MSE is 0.395 versus 0.404 for S²TS-LLM across four horizons; gains are not universal.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.