Zobrazeno 1 - 10
of 192
pro vyhledávání: '"li, Hanchen"'
Because of reinforcement learning's (RL) ability to automatically create more adaptive controlling logics beyond the hand-crafted heuristics, numerous effort has been made to apply RL to congestion control (CC) design for real time video communicatio
Externí odkaz:
http://arxiv.org/abs/2411.06742
Autor:
Liu, Yuhan, Huang, Yuyang, Yao, Jiayi, Gu, Zhuohan, Du, Kuntai, Li, Hanchen, Cheng, Yihua, Jiang, Junchen, Lu, Shan, Musuvathi, Madan, Choukse, Esha
Large Language Models (LLMs) are increasingly employed in complex workflows, where different LLMs and fine-tuned variants collaboratively address complex tasks. However, these systems face significant inefficiencies due to redundant context processin
Externí odkaz:
http://arxiv.org/abs/2411.02820
Autor:
Yao, Jiayi, Li, Hanchen, Liu, Yuhan, Ray, Siddhant, Cheng, Yihua, Zhang, Qizheng, Du, Kuntai, Lu, Shan, Jiang, Junchen
Large language models (LLMs) often incorporate multiple text chunks in their inputs to provide the necessary contexts. To speed up the prefill of the long LLM inputs, one can pre-compute the KV cache of a text and re-use the KV cache when the context
Externí odkaz:
http://arxiv.org/abs/2405.16444
Autor:
Li, Hanchen, Zhu, Chaofeng
In this paper, we prove the stability theorems for the isotropic perturbations of maximal isotropic subspaces in symplectic Banach spaces. Then we prove a stability theorem for the mod $2$ dimensions of kernel of skew-adjoint linear Fredholm relation
Externí odkaz:
http://arxiv.org/abs/2404.08445
To render each generated token in real-time for users, the Large Language Model (LLM) server generates tokens one by one and streams each token (or group of a few tokens) through the network to the user right after generation, which we refer to as LL
Externí odkaz:
http://arxiv.org/abs/2401.12961
For video or web services, it is crucial to measure user-perceived quality of experience (QoE) at scale under various video quality or page loading delays. However, fast QoE measurements remain challenging as they must elicit subjective assessment fr
Externí odkaz:
http://arxiv.org/abs/2311.06698
Autor:
Liu, Yuhan, Li, Hanchen, Cheng, Yihua, Ray, Siddhant, Huang, Yuyang, Zhang, Qizheng, Du, Kuntai, Yao, Jiayi, Lu, Shan, Ananthanarayanan, Ganesh, Maire, Michael, Hoffmann, Henry, Holtzman, Ari, Jiang, Junchen
As large language models (LLMs) take on complex tasks, their inputs are supplemented with longer contexts that incorporate domain knowledge. Yet using long contexts is challenging, as nothing can be generated until the whole context is processed by t
Externí odkaz:
http://arxiv.org/abs/2310.07240
Autor:
Cheng, Yihua, Zhang, Ziyi, Li, Hanchen, Arapin, Anton, Zhang, Yue, Zhang, Qizheng, Liu, Yuhan, Zhang, Xu, Yan, Francis Y., Mazumdar, Amrita, Feamster, Nick, Jiang, Junchen
In real-time video communication, retransmitting lost packets over high-latency networks is not viable due to strict latency requirements. To counter packet losses without retransmission, two primary strategies are employed -- encoder-based forward e
Externí odkaz:
http://arxiv.org/abs/2305.12333
Autor:
Cheng, Yihua, Arapin, Anton, Zhang, Ziyi, Zhang, Qizheng, Li, Hanchen, Feamster, Nick, Jiang, Junchen
Across many real-time video applications, we see a growing need (especially in long delays and dynamic bandwidth) to allow clients to decode each frame once any (non-empty) subset of its packets is received and improve quality with each new packet. W
Externí odkaz:
http://arxiv.org/abs/2210.16639
Autor:
Li, Hanchen, Ding, Ruochen, Su, Wen, Lin, Xinxing, Guan, Sumin, Ye, Qingping, Zheng, Zhimei, Wang, Jiaqiang
Publikováno v:
In Energy Conversion and Management 1 November 2024 319
