7 papers · 1 filter
ChampionKV: Head-Wise Adaptive Budget Allocation for KV Cache Compression
Yuechuan Chen, Yanda Guo, Xiaoxi Liu +4
Some observation-window-based KV cache compressors use prompt attention to identify important tokens but allocate uniform or predefined budgets across heads, overlooking input-depe…
SEDformer: Event-Synchronous Spiking Transformers for Irregular Telemetry Time Series Forecasting
Ziyu Zhou, Yuchen Fang, Weilin Ruan +3
Telemetry streams from large-scale Internet-connected systems (e.g., IoT deployments and online platforms) naturally form an irregular multivariate time series (IMTS) whose accurat…
TS-Memory: Plug-and-Play Memory for Time Series Foundation Models
Sisuo Lyu, Siru Zhong, Tiegang Chen +6
Time Series Foundation Models (TSFMs) achieve strong zero-shot forecasting through large-scale pre-training, but adapting them to downstream domains under distribution shift remain…
A Game-Theoretic Spatio-Temporal Reinforcement Learning Framework for Collaborative Public Resource Allocation
Songxin Lei, Qiongyan Wang, Yanchen Zhu +6
Public resource allocation involves distributing resources, including urban infrastructure, energy, and transportation, which are typically limited in capacity, to meet social dema…
RAST: A Retrieval Augmented Spatio-Temporal Framework for Traffic Prediction
Weilin Ruan, Xilin Dang, Ziyu Zhou +2
Traffic prediction is a cornerstone of modern intelligent transportation systems and a critical task in spatio-temporal forecasting. Although advanced Spatio-temporal Graph Neural…
OccamVTS: Distilling Vision Models to 1% Parameters for Time Series Forecasting
Sisuo Lyu, Siru Zhong, Weilin Ruan +4
Time series forecasting is fundamental to diverse applications, with recent approaches leverage large vision models (LVMs) to capture temporal patterns through visual representatio…
ST-LoRA: Low-rank Adaptation for Spatio-Temporal Forecasting
Weilin Ruan, Wei Chen, Xilin Dang +4
Spatio-temporal forecasting is essential for understanding future dynamics within real-world systems by leveraging historical data from multiple locations. Existing methods often p…