6 papers
RelaxKV: Recomputation Guided by the Query with Sparse Context Attention for Efficient KV Cache Reuse
Ruoling Qi, Yirui Liu, Xuaner Wu +5
Cross-request KV caching reduces the prefill cost of Retrieval-Augmented Generation (RAG), but conventional prefix caching severely limits cache reuse across requests. Position-Ind…
Just Let Linear States Forget the Distant Past: Prefix Caching via Suffix Replay for Hybrid LLMs
Yirui Liu, Ruoling Qi, Xuaner Wu +6
Hybrid LLMs interleave full-attention layers with linear-attention layers to reduce long-context inference cost, but this structure complicates prefix caching. Full-attention KV ca…
REVISE: Validity-Guided Recovery for Online Revisions in Agent Workflows
Ruoling Qi, Xuaner Wu, Penghang Liu +2
Agent revisions expose a fundamental correctness--efficiency trade-off during concurrent execution. Discarding ongoing work preserves latest-version correctness but wastes progress…
Tail-Replay: Escaping the Curse of Linear Attention in Prefix Caching for Hybrid LLMs
Yirui Liu, Ruoling Qi, Xuaner Wu +2
Hybrid large language models interleave full-attention layers with linear-attention layers to reduce the cost of long-context inference. This structure complicates prefix caching:…
LinearKV: One Cached State Suffices for Position-Independent Caching in Hybrid LLMs
Yirui Liu, Ruoling Qi, Longwen Wang +5
LLM serving is increasingly accelerated by position-independent caching (PIC). Existing PIC methods, however, are built for full-attention models, where a token-indexed KV cache un…
Swift-SVD: Theoretical Optimality Meets Practical Efficiency in Low-Rank LLM Compression
Ruoling Qi, Yirui Liu, Xuaner Wu +6
The deployment of Large Language Models is constrained by the memory and bandwidth demands of static weights and dynamic Key-Value cache. SVD-based compression provides a hardware-…