Zihao Xuan

Zihao Xuan, Jia Chen, Yewen Li et al.

In this paper, we propose FusionCIM, an operator-fusion-driven compute-in-memory (CIM) accelerator architecture for efficient and scalable LLM inference, with three key innovations: (1) a hybrid CIM pipeline architecture that maps QKT computation on inner-product-based CIM (IP-CIM) and PV aggregation on outer-product-based CIM (OP-CIM) for efficient matrix multiplications fusion; (2) a QO-stationary dataflow that eliminates repeated KV loading in CIM and K-matrix access in buffer under transpose fusion, significantly improving data reuse on chip; and (3) a pattern-aware online-softmax mechanism that exploits distribution regularities of attention scores to reduce exponential rescaling overhead for non-linear fusion. Experimental results on LLaMA-3 model show that FusionCIM achieves up to 3.86x energy saving, and 1.98x speedup compared with prior SOTA CIM-based designs with 29.4 TOPS/W energy efficiency at the system level.

Chao Han, Yijuan Liang, Zihao Xuan et al.

The deployment of large language models (LLMs) in real-world applications is increasingly limited by their high inference cost. While recent advances in dynamic token-level computation allocation attempt to improve efficiency by selectively activating model components per token, existing methods rely on greedy routing--a myopic execute-or-skip mechanism that often leads to irreversible information loss and suboptimal token selection. This paper introduces informed routing, a new paradigm that proactively addresses these issues. The key insight is to assess not only a token's immediate importance but also its recoverability, i.e., how well its transformation can be approximated. To this end, we propose the Lightweight Feature Forecaster (LFF), a small predictive module that estimates a unit's output before routing decisions are made. This enables a flexible execute-or-approximate policy that preserves model fidelity while drastically reducing computation. Extensive experiments on both language modeling and reasoning tasks show that informed routing achieves state-of-the-art efficiency-performance trade-offs across multiple sparsity levels. Notably, even without final LoRA fine-tuning, our method matches or surpasses strong baselines that require full fine-tuning, all while reducing training time by over 50%. The code is available at: https://github.com/EIT-NLP/informed-routing

Zihao Xuan, Yuxuan Yang, Wei Xuan et al.

In this paper, we further explore the potential of analog in-memory computing (AiMC) and introduce an innovative artificial intelligence (AI) accelerator architecture named YOCO, featuring three key proposals: (1) YOCO proposes a novel 8-bit in-situ multiply arithmetic (IMA) achieving 123.8 TOPS/W energy-efficiency and 34.9 TOPS throughput through efficient charge-domain computation and timedomain accumulation mechanism. (2) YOCO employs a hybrid ReRAM-SRAM memory structure to balance computational efficiency and storage density. (3) YOCO tailors an IMC-friendly attention computing flow with an efficient pipeline to accelerate the inference of transformer-based AI models. Compared to three SOTA baselines, YOCO on average improves energy efficiency by up to 3.9x-19.9x and throughput by up to 6.8x-33.6x across 10 CNN/transformer models.