Moo-Woong JEONG †a) , Nonmember, Tae-Won BAN †b) , Member, and Bang Chul JUNG † †c) , Nonmember SUMMARY In this paper, we investigate a user and antenna joint selection problem in multiuser large-scale MIMO downlink networks, where a BS with N transmit antennas serves K users, and N is much larger than K. The BS activates only S(S ≤ N) antennas for data transmission to reduce hardware cost and computation complexity, and selects the set of users to which data is to be transmitted by maximizing the sum-rate. The optimal user and antenna joint selection scheme based on exhaustive search causes considerable computation complexity. Thus, we propose a new joint selection algorithm with low complexity and analyze the performance of the proposed scheme in terms of sum-rate and complexity. When S = 7, N = 10, K = 5, and SNR=10 dB, the sum-rate of the proposed scheme is 5.1% lower than that of the optimal scheme, while the computation complexity of the proposed scheme is reduced by 99.0% compared to that of the optimal scheme.

No abstract

Recently, Large Multi-modal Models (LMMs) have demonstrated their ability to understand the visual contents of images given the instructions regarding the images. Built upon the Large Language Models (LLMs), LMMs also inherit their abilities and characteristics such as in-context learning where a coherent sequence of images and texts are given as the input prompt. However, we identify a new limitation of off-the-shelf LMMs where a small fraction of incoherent images or text descriptions mislead LMMs to only generate biased output about the hijacked context, not the originally intended context. To address this, we propose a pre-filtering method that removes irrelevant contexts via GPT-4V, based on its robustness towards distribution shift within the contexts. We further investigate whether replacing the hijacked visual and textual contexts with the correlated ones via GPT-4V and text-to-image models can help yield coherent responses.

No abstract

No abstract