REVIEW 9 cited by
Are Language Models Actually Useful for Time Series Forecasting?
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
Signed reviews
read the original abstract
Large language models (LLMs) are being applied to time series forecasting. But are language models actually useful for time series? In a series of ablation studies on three recent and popular LLM-based time series forecasting methods, we find that removing the LLM component or replacing it with a basic attention layer does not degrade forecasting performance -- in most cases, the results even improve! We also find that despite their significant computational cost, pretrained LLMs do no better than models trained from scratch, do not represent the sequential dependencies in time series, and do not assist in few-shot settings. Additionally, we explore time series encoders and find that patching and attention structures perform similarly to LLM-based forecasters.
Forward citations
Cited by 9 Pith papers
-
CSI-4CAST: A Hybrid Deep Learning Model for CSI Prediction with Comprehensive Robustness and Generalization Testing
A hybrid CNN-ShuffleNet-Transformer model claims the best accuracy-efficiency trade-off on a new 3,060-scenario CSI prediction benchmark, but FDD gains and robustness claims rest on point estimates without error bars.
-
Integrating Traditional Technical Analysis with AI: A Multi-Agent LLM-Based Approach to Stock Market Forecasting
An LLM-based multi-agent system that identifies Elliott Wave patterns achieved 44-89% directional accuracy on six US stocks, with deep reinforcement learning backtesting improving accuracy on most but not all test cases.
-
TempoGPT: Enhancing Time Series Reasoning via Quantizing Embedding
Converting time series into discrete tokens via a VQ-VAE codebook and feeding them through a shared embedding layer with text improves a language model's accuracy on a synthetic circuit-based time series reasoning benchmark.
-
Learning Latent Spaces for Domain Generalization in Time Series Forecasting
LTG decomposes series into trend and seasonal parts, learns latent factors with a conditional beta-VAE and domain regularization, and feeds them to a forecaster to improve generalization to unseen domains.
-
Financial Fine-tuning a Large Time Series Model
Fine-tuning TimesFM on log-transformed financial price data improves directional accuracy and mock-trading Sharpe ratios over the base model.
-
Adversarial Vulnerabilities in Large Language Models for Time Series Forecasting
A simple gradient-free 'directional gradient approximation' attack makes LLM time series forecasters degrade more than equivalent random noise, across GPT-3.5, GPT-4, LLaMa, Mistral, TimeGPT, and TimeLLM.
-
Fine-Tuning Pre-trained Large Time Series Models for Prediction of Wind Turbine SCADA Data
Fine-tuning the pre-trained Timer model on a single wind turbine's SCADA data yields the best prediction accuracy across the whole wind farm, while the model does not always beat simpler baselines on larger datasets.
-
ElliottAgents: A Natural Language-Driven Multi-Agent System for Stock Market Analysis and Prediction
A natural-language multi-agent system combines Elliott Wave pattern detection with LLM dialogue and DRL backtesting to generate stock predictions, supported only by small selected experiments.
-
Revisiting Data Analysis with Pre-trained Foundation Models
A comprehensive review of using pre-trained foundation models for data analysis, with a taxonomy of tasks, challenges, and a research roadmap.
Discussion (0). Continue with ORCID to comment.