The research achievement "SALUT: Signed-Digit Attention with Unified LUT-Based Compute for Ternary LLM Inference" by Professor Wang Chao's team from the School of Optical and Electronic Information at our university has been accepted by the IEEE International Symposium on High-Performance Computer Architecture (HPCA 2027), a leading international conference in the field of computer architecture. This work proposes a unified look-up-table (LUT)-based computing architecture for ternary large language models (LLMs) that enables the compressed attention cache to participate directly in computation, achieves hardware resource sharing between linear-layer and attention computations, and offers a novel hardware solution that balances storage compression and computational efficiency for low-bit LLM inference.