Zhongruo Wang — accepted AI-conference papers

2026

GmNet: Revisiting Gating Mechanisms From A Frequency View

Yifan Wang, Xu Ma, Yitian Zhang, Yizhou Wang, Zhongruo Wang, Sung-Cheol Kim +3

ICLR 2026poster

Lightweight neural networks, essential for on-device applications, often suffer from a low-frequency bias due to their constrained capacity and depth. This limits their ability to capture the fine-grained, high-frequency details (e.g., textures, edges) that are crucial for complex computer vision t…

Cited by 0 Source ↗Scholar ↗

2026

SFT Doesn’t Always Hurt General Capabilities: Revisiting Domain-Specific Fine-Tuning in LLMs

Jiacheng Lin, Zhongruo Wang, Kun Qian, Tian Wang, Arvind Srinivasan, Hansi Zeng +11

ICLR 2026poster

Supervised Fine-Tuning (SFT) on domain-specific datasets is a common approach to adapt Large Language Models (LLMs) to specialized tasks but is often believed to degrade their general capabilities. In this work, we revisit this trade-off and present both empirical and theoretical insights. First, we…

Cited by 0 Source ↗Scholar ↗