Agent applications are evolving into workflows composed of large language model (LLM) calls and tool invocations, termed modules. Existing distributed agent systems typically adopt a disaggregated architecture, deploying different modules on separate devices. While this enables pipeline parallelism, imbalanced module l...
Kai-Bin Guo, Peng-Tu Li, Zi-Cong Hong et al.· 0 citations
Federated learning (FL) is a privacy-preserving machine learning (ML) paradigm that can learn models from distributed datasets owned by mobile terminals (MTs). However, ML models usually contain bias on some user groups/sensitive attributes (e.g., gender and race), which poses additional challenges on FL in real-world...
Kang Wei, Xin-Nan Yuan, Zi-Cong Hong et al.· IEEE Transactions on Mobile...· 0 citations
SteerQuant, a 4-bit quantization framework for WAMs that steers errors toward computations with less influence on final actions, is introduced and Rudder, a 4-bit inference engine for WAMs that fuses scaling and output compensation into low-bit kernels is developed.
Yun-Han Wang, Hao-Dong Wang, Zhi-Ming Liu et al.· arXiv.org· 0 citations
Pilot, an efficient engine that reduces sparse inference overhead through lightweight scoring and cross-step reuse of token selections, is introduced, an efficient engine that reduces sparse inference overhead through lightweight scoring and cross-step reuse of token selections.
Xin-Ling Xie, Hao-Dong Wang, Jia-Zhi Mi et al.· arXiv.org· 0 citations
Generative world models aim to predict future states conditioned on actions, where action controllability is fundamental for reliable dynamics modeling. While recent efforts leverage simulator-generated data to enhance this capability, existing training pipelines face two fundamental limitations. First, static offline...
Yi-Kun Miao, Fang-Qi Zhu, Quan-Xin Shou et al.· 2 citations
K AIROX introduces a Live Pipeline designed to prefetch neurons by predicting next-layer activation patterns, a mechanism that dynamically redistributes neurons between the GPU and CPU based on activation patterns, and a Temporal Activation Momentum cache policy to prioritize neurons with sustained utility while minimi...
Yapeng Jiang, Minghao Gan, Zi-Cong Hong et al.· 0 citations
This work introduces RedFlow, an offline post-training method for flow-matching VLA policies that requires neither external HIL corrections nor online data collection during post-training, and improves average success across four LIBERO suites.