Research map / Time series / Long-term forecasting
L-Drive: Beyond a Single Mapping—Latent Context Drives Time Series Forecasting
Fan Zhang, Shijun Chen, Hua Wang
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 L-Context generator (ablation) | L-Drive · ETTh2 | MSE: 0.375 -> 0.367 | Improved |
| add L-Context generator (ablation) | L-Drive · ECL | MSE: 0.177 -> 0.169 | Improved |
| add gating on increment representations (ablation) | L-Drive · ECL | MSE: 0.176 -> 0.169 | Improved |
| use structured L-Context instead of random learnable context variables (ablation) | L-Drive · ETTh2 | MSE: 0.382 -> 0.367 | Improved |
| use intra-patch relative positional bases instead of none (ablation) | L-Drive · ETTm2 | MSE: 0.277 -> 0.273 | Improved |
| use patch-shared relative positional bases instead of global positional representation (ablation) | L-Drive · ETTh1 | MSE: 0.434 -> 0.427 | Improved |
| use patch-shared relative positional bases instead of global positional representation (ablation) | L-Drive · ETTm2 | MSE: 0.273 -> 0.273 | No clear effect |
| inject L-Context into DLinear | DLinear · ECL | MSE: 0.212 -> 0.198 | Improved |
| inject L-Context into PatchTST | PatchTST · ECL | MSE: 0.205 -> 0.184 | Improved |
| inject L-Context into iTransformer | iTransformer · ECL | MSE: 0.178 -> 0.176 | Improved |
| inject L-Context into DLinear | DLinear · Weather | MSE: 0.265 -> 0.265 | No clear effect |
| use L-Drive instead of Direct-Mapping baselines around change events | L-Drive vs TimeBridge and iTransformer · ETTh1 lag error | TailAUC: 119.98 -> 95.94 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 3 of 3 on most shared results.
- TimeBridge — TimeBridge: Non-Stationarity Matters for Long-term Time Series Forecasting (ICML 2025) · better on 4 of 5 shared results
- iTransformer — iTransformer: Inverted Transformers Are Effective for Time Series Forecasting (ICLR 2024) · better on 2 of 2 shared results
- TimeXer — TimeXer: Empowering Transformers for Time Series Forecasting with Exogenous Variables (NeurIPS 2024) · better on 1 of 1 shared result
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
- Rethinking Post-Training Recipes for Multimodal Time-Series Forecasting (NeurIPS 2026 · similarity 0.88)
- TimeBase: The Power of Minimalism in Efficient Long-term Time Series Forecasting (ICML 2025 · similarity 0.87)
- ARM: Refining Multivariate Forecasting with Adaptive Temporal-Contextual Learning (ICLR 2024 · similarity 0.87)
- Dirichlet-Guided Group Forecasting for Alleviating Over-smoothing in Time Series Forecasting (NeurIPS 2026 · similarity 0.86)
- TimeXL: Explainable Multi-modal Time Series Prediction with LLM-in-the-Loop (NeurIPS 2025 · similarity 0.86)
- Learning Pattern-Specific Experts for Time Series Forecasting Under Patch-level Distribution Shift (NeurIPS 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.
