Deep TPC: Temporal-Prior Conditioning for Time Series Forecasting
On the map: LLM reasoning → Time series reasoning
Abstract (arXiv)
LLM-for-time series (TS) methods typically treat time shallowly, injecting positional or prompt-based cues once at the input of a largely frozen decoder, which limits temporal reasoning as this information degrades through the layers. We introduce Temporal-Prior Conditioning (TPC), which elevates time to a first-class modality that conditions the model at multiple depths. TPC attaches a small set of learnable time series tokens to the patch stream; at selected layers these tokens cross-attend to temporal embeddings derived from compact, human-readable temporal descriptors encoded by the same frozen LLM, then feed temporal context back via self-attention. This disentangles time series signal and temporal information while maintaining a low parameter budget. We show that by training only the cross-attention modules and explicitly disentangling time series signal and temporal information, TPC consistently outperforms both full fine-tuning and shallow conditioning strategies, achieving state-of-the-art performance in long-term forecasting across diverse datasets. Code available at: https://github.com/fil-mp/Deep_tpc
Key facts. No citing papers in the map yet (data through 7 Oct 2026). Placed in Time series reasoning, one of 17 directions in LLM reasoning. Builds on 18 of its 21 references that are in the map. First posted on arXiv 18 Feb 2026.
Builds on
18 of 21 references.
References · 18
- 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Modelscited by 12,302
- 2023BLIP-2: Bootstrapping Language-Image Pre-training with Frozen Image Encoders and Large Language Modelscited by 6,069
- 2022A Time Series is Worth 64 Words: Long-term Forecasting with Transformerscited by 1,981
- 2022Are Transformers Effective for Time Series Forecasting?cited by 1,400
- 2022TimesNet: Temporal 2D-Variation Modeling for General Time Series Analysiscited by 1,030
- 2024Chronos: Learning the Language of Time Seriescited by 712
- 2023Time-LLM: Time Series Forecasting by Reprogramming Large Language Modelscited by 680
- 2021Perceiver IO: A General Architecture for Structured Inputs & Outputscited by 616
- 2023One Fits All:Power General Time Series Analysis by Pretrained LMcited by 615
- 2023TEST: Text Prototype Aligned Embedding to Activate LLM's Ability for Time Seriescited by 165
- 2023A Survey on Graph Neural Networks for Time Series: Forecasting, Classification, Imputation, and Anomaly Detectioncited by 145
- 2023UniTime: A Language-Empowered Unified Model for Cross-Domain Time Series Forecastingcited by 139
6 more
- 2024From News to Forecast: Integrating Event Analysis in LLM-Based Time Series Forecasting with Reflectioncited by 93
- 2022Domain Adversarial Spatial-Temporal Network: A Transferable Framework for Short-term Traffic Forecasting across Citiescited by 38
- 2023SaDI: A Self-adaptive Decomposed Interpretable Framework for Electric Load Forecasting under Extreme Eventscited by 21
- 2024CTPD: Cross-Modal Temporal Pattern Discovery for Enhanced Multimodal Electronic Health Records Analysiscited by 6
- 2025DeepMLF: Multimodal language model with learnable tokens for deep fusion in sentiment analysiscited by 3
- 2024VITRO: Vocabulary Inversion for Time-series Representation Optimizationcited by 1
Built on by
Not cited yet.
Similar, not cited
Close in wording, no citation either way
- 2026MemCast: Memory-Driven Time Series Forecasting with Experience-Conditioned Reasoning
- 2026Enhancing few-shot time series forecasting with LLM-guided diffusion
- 2025Time Series Forecasting via Reasoning: A Slow-Thinking Approach with Reinforcement Fine-Tuned LLMs
- 2025TsLLM: Augmenting LLMs for General Time Series Understanding and Prediction
- 2025Small but Mighty: Enhancing Time Series Forecasting with Lightweight LLMs
- 2026Rethinking Post-Training Recipes for Multimodal Time-Series Forecasting
- 2023LLM4TS: Aligning Pre-Trained LLMs as Data-Efficient Time-Series Forecasters
- 2025Adapting LLMs to Time Series Forecasting via Temporal Heterogeneity Modeling and Representation Alignment