@inproceedings{fan-etal-2025-position,
    title = "Position-Aware Depth Decay Decoding ($D^3$): Boosting Large Language Model Inference Efficiency",
    author = "Fan, Siqi  and
      Fang, Xuezhi  and
      Xing, Xingrun  and
      Han, Peng  and
      Shang, Shuo  and
      Wang, Yequan",
    editor = "Che, Wanxiang  and
      Nabende, Joyce  and
      Shutova, Ekaterina  and
      Pilehvar, Mohammad Taher",
    booktitle = "Findings of the Association for Computational Linguistics: ACL 2025",
    month = jul,
    year = "2025",
    address = "Vienna, Austria",
    publisher = "Association for Computational Linguistics",
    url = "https://preview.aclanthology.org/acl25-workshop-ingestion/2025.findings-acl.154/",
    pages = "2990--3001",
    ISBN = "979-8-89176-256-5",
    abstract = "Due to the large number of parameters, the inference phase of Large Language Models (LLMs) is resource-intensive. Unlike traditional model compression, which needs retraining, recent dynamic computation methods show that not all components are required for inference, enabling a training-free pipeline.In this paper, we focus on the dynamic depth of LLM generation. A token-position aware layer skipping framework is proposed to save 1.5x times operations efficiently while maintaining performance.We first observed that tokens predicted later have lower perplexity and thus require less computation. Then, we propose a training-free algorithm called Position-Aware \textbf{D}epth \textbf{D}ecay \textbf{D}ecoding (), which leverages a power-law decay function, $\left\lfloor L \times (\alpha^i) \right\rfloor$, to determine the number of layers to retain when generating token $T_i$. Remarkably, without any retraining, the achieves success across a wide range of generation tasks for the first time.Experiments on large language models (the Llama) with $7 \sim 70$ billion parameters show that can achieve an average 1.5x speedup compared with the full-inference pipeline while maintaining comparable performance with nearly no performance drop ($<1\%$) on the GSM8K and BBH benchmarks."
}