Li Li

Xuyang Wei, Chunlin Tian, Li Li

Effective instruction fine-tuning on diverse image-text datasets is crucial for developing a versatile Multimodal Large Language Model (MLLM), where dataset composition dictates the model's adaptability across multimodal tasks. However, complex datasets often contain inherent conflicts -- stemming from modality-specific optimization objectives -- and latent commonalities that enable cross-task transfer, which most existing approaches handle separately. To bridge this gap, we introduce AsymLoRA, a parameter-efficient tuning framework that unifies knowledge modularization and cross-modal coordination via asymmetric LoRA: task-specific low-rank projections (matrix B) that preserve distinct adaptation pathways for conflicting objectives, and a shared projection (matrix A) that consolidates cross-modal commonalities. Extensive evaluations demonstrate that AsymLoRA consistently surpasses both vanilla LoRA, which captures only commonalities, and LoRA-MoE, which focuses solely on conflicts, achieving superior model performance and system efficiency across diverse benchmarks.\href{Code}{https://github.com/Clin0212/HydraLoRA/blob/main/MLLM-HydraLoRA/README.md}.

Kamesh Namuduri, Li Li, Mahadevan Gomathisankaran et al.

This paper presents the chief security officer (CSO) problem, defines its scope, and investigates several important research questions related within the scope. The CSO problem is defined based on the concept of secrecy capacity of wireless communication channels. It is also related to the chief Estimation/Executive Officer (CEO) problem that has been well studied in information theory. The CSO problem consists of a CSO, several agents capable of having two-way communication with the CSO, and a group of eavesdroppers. There are two scenarios in the CSO problem; one in which agents are not allowed to cooperate with one another and the other in which agents are allowed to cooperate with one another. While there are several research questions relevant to the CSO problem, this paper focusses on the following and provides answers: (1) How much information can be exchanged back and forth between the CSO and the agents without leaking any information to the eavesdroppers? (2) What is the power allocation strategy that the CSO needs to follow so as to maximize the secrecy capacity? (3) How can agents cooperate with one another in order to increase the overall secrecy capacity?.