paper
Beyond Uniform Tokens: Adaptive Compression for Time Series Language Models
paperactiveprovisional
beyond-uniform-tokens-adaptive-compression-for-time-series-language-models-6efc6599·1 events·first seen 5d agoAliases: Beyond Uniform Tokens: Adaptive Compression for Time Series Language Models
More like this (12)
Adaptive Multi-Resolution Procedural Knowledge Compression for Large Language ModelsCLP: Collocation-Length Prediction for Zero-Loss Adaptive Multi-Token InferencePlanning-aligned Token Compression for Long-Context Autonomous Drivingmultivariate time series representation learningBeyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language ModelsSelf-Augmenting Retrieval for Diffusion Language Modelsvisual-token compressionTransformer Language ModelsLESS: Mutual-Stability Sampling for Diffusion Language ModelsTime Series Foundation ModelsExploring Adversarial Robustness and Safety Alignment in Multilingual Multi-Modal Large Language Modelsencoder-only language models
Recent events (1)
Adaptive asymmetric token compression accelerates time series language models up to 7.68×
A new arXiv preprint proposes an adaptive token budgeting framework for time series (TS) language models that compresses TS tokens using frequency-domain structure and progressively prunes prompt tokens across model layers. The authors demonstrate up to 7.68× inference acceleration with performance improvements in 78% of evaluated settings across forecasting, classification, imputation, and anomaly detection tasks. The work is motivated by the observation that TS tokens have uneven spectral contributions and prompt-token influence attenuates with model depth, making uniform token processing wasteful.