Hao Shi
2 indexed papers
Publications per year
Top categories
Frequent co-authors
Research Timeline
This paper demonstrates that adversarial perturbations possess a low-rank structure, and proposes a two-step method to leverage this property to significantly improve the efficiency and effectiveness of black-box adversarial attacks.
This paper proposes a training-free framework called ReasonAlloc to mitigate inference bottlenecks in large language models by recasting decoding-time key-value compression as a hierarchical budget allocation problem.
Papers
ReasonAlloc: Hierarchical Decoding-Time KV Cache Budget Allocation for Reasoning Models
Wenhao Liu, Hao Shi, Yunhe Li, Weizhi Fei +6 more
This paper proposes a training-free framework called ReasonAlloc to mitigate inference bottlenecks in large language models by recasting decoding-time key-value compression as a hierarchical budget al…