README.md

August 11, 2026 Ā· View on GitHub

Efficient Test-Time Scaling for LLM-based Time Series Forecasting

Architecture

SCALER architecture

Stage I: Predict a coarse future shape using a lightweight global forecaster and compare with coarse version of the target.

Stage II: Use the coarse forecast as a global anchor and let a frozen LLM iteratively refine local details over K steps.

ā†’ Decode the refined tokens back into the original resolution to obtain the final forecast.

Visualization

SCALER visualization

SCALER faithfully reproduces peaks, troughs, and amplitude patterns, yielding structurally coherent predictions beyond aggregate error improvements.

Requirements

Use python 3.11 from MiniConda

  • torch==2.2.2
  • accelerate==0.28.0
  • einops==0.7.0
  • matplotlib==3.7.0
  • numpy==1.23.5
  • pandas==1.5.3
  • scikit_learn==1.2.2
  • scipy==1.12.0
  • tqdm==4.65.0
  • peft==0.4.0
  • transformers==4.31.0
  • deepspeed==0.14.0
  • sentencepiece==0.2.0

To install all dependencies:

pip install -r requirements.txt

Datasets

You can access the well pre-processed datasets from [Google Drive], then place the downloaded contents under ./dataset