Research map / Time series / Foundation models for time series
Invariant Representation Learning for Source-Free Time Series Forecasting with LLM-Centric Proxy Denoising
Kangjia Yan, Chenxi Liu, Hao Miao, Xinle Wu, Yan Zhao, Chenjuan Guo, Bin Yang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add knowledge distillation from the denoised LLM to the target model (ablation) | TimeID · ETTh1 -> ETTh2 | MSE / MAE: removal raises MSE 17.86% and MAE 18.34% | Improved |
| add invariant disentangled feature learning (IDFL) (ablation) | TimeID · ETTh1 -> Traffic | MSE / MAE: removal raises MSE 20.13% and MAE 26.91% | Improved |
| add proxy denoising of LLM predictions (ablation) | TimeID · ETTh1 target datasets | MSE / MAE | Improved |
| add LLM guidance (OFA backbone) (ablation) | TimeID · ETTh1 target datasets | MSE / MAE | Improved |
| add gradient-level invariant block (ablation) | TimeID · ETTh1 target datasets | MSE / MAE | Improved |
| increase target training data from 5% to 40% (ablation) | TimeID · ETTh1 -> ETTh2 and ETTh1 -> ETTm2 | MSE / MAE | Improved |
| use TimeID instead of Time-LLM and OFA | source-free forecasting · 18 tasks from ETTh1 | average MSE reduction vs Time-LLM: 4.98% MSE reduction vs Time-LLM, 4.39% vs OFA | Improved |
| use TimeID instead of pretrained time series foundation models | 30% few-shot · ETTh1 -> Traffic | MSE: 0.452 vs MOIRAI 0.368 | Hurt |
| use TimeID instead of foundation models | 30% few-shot · ETTh1 -> ETTh2 and Weather | MSE: 3.4% and 11.1% lower MSE than best foundation model | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- Time-LLM — Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Time-Series Forecasting for Out-of-Distribution Generalization Using Invariant Learning (ICML 2024 · similarity 0.88)
- TimeEmb: A Lightweight Static-Dynamic Disentanglement Framework for Time Series Forecasting (NeurIPS 2025 · similarity 0.88)
- Time-LLM: Time Series Forecasting by Reprogramming Large Language Models (ICLR 2024 · similarity 0.87)
- TimePFN: Effective Multivariate Time Series Forecasting with Synthetic Data (AAAI 2025 · similarity 0.87)
- Towards a General Time Series Forecasting Model with Unified Representation and Adaptive Transfer (ICML 2025 · similarity 0.86)
- Unlocking the Power of Patch: Patch-Based MLP for Long-Term Time Series Forecasting (AAAI 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
