paper-with-me

홈 › Papers

Dynamic Vocabulary Pruning in Early-Exit LLMs

2024-10-24 · Jort Vincenti, Karim Abdel Sadek, Joan Velja, Matteo Nulli, Metod Jazbec

Increasing the size of large language models (LLMs) has been shown to lead to better performance. However, this comes at the cost of slower and more expensive inference. Early-exiting is a promising approach for improving the efficiency of LLM inference by enabling next token prediction at intermediate layers. Yet, the large vocabulary size in modern LLMs makes the confidence estimation required for exit decisions computationally expensive, diminishing the efficiency gains. To address this, we propose dynamically pruning the vocabulary at test time for each token. Specifically, the vocabulary is pruned at one of the initial layers, and the smaller vocabulary is then used throughout the rest of the forward pass. Our experiments demonstrate that such post-hoc dynamic vocabulary pruning improves the efficiency of confidence estimation in early-exit LLMs while maintaining competitive performance.

📄 PDF Abstract BibTeX arXiv:2410.18952

Code (1)

matteonulli/vocabulary_pruning 공식 구현 pytorch

Methods 이 논문이 사용한 방법론

Pruning 설명 없음

Similar Papers 제목 키워드 기반

Balanced and Elastic End-to-end Training of Dynamic LLMs

2025-05-20 · Mohamed Wahib, Muhammed Abdullah Soyturk, Didem Unat

To reduce computational and memory costs in Large Language Models (LLMs), dynamic workload reduction schemes like Mixture of Experts (MoEs), parameter pruning, layer freezing, sparse attention, early token exit, and Mixt…

GPUMixture-of-Experts

HiDrop: Hierarchical Vision Token Reduction in MLLMs via Late Injection, Concave Pyramid Pruning, and Early Exit

2026-02-27 · Hao Wu, Yingqi Fan, Jinyang Dai, Junlong Tong 외 arxiv

The quadratic computational cost of processing vision tokens in Multimodal Large Language Models (MLLMs) hinders their widespread adoption. While progressive vision token pruning offers a promising solution, current meth…

Pruning Early Exit Networks

2022-07-08 · Alperen Görmez, Erdem Koyuncu

Deep learning models that perform well often have high computational costs. In this paper, we combine two approaches that try to reduce the computational cost while keeping the model performance high: pruning and early e…

Dynamic Vocabulary Pruning: Stable LLM-RL by Taming the Tail

2025-12-28 · Yingru Li, Jiawei Xu, Jiacai Liu, Yuxuan Tong 외 arxiv

Reinforcement Learning (RL) for Large Language Models (LLMs) faces a fundamental tension: the numerical divergence between high-throughput inference engines and numerically precise training engines. Although these system…

Reinforcement Learning

A Comparative Study of CNN Optimization Methods for Edge AI: Exploring the Role of Early Exits

2026-04-16 · Nekane Fernandez, Ivan Valdes, Steven Van Vaerenbergh, Idoia de la Iglesia 외 arxiv

Deploying deep neural networks on edge devices requires balancing accuracy, latency, and resource constraints under realistic execution conditions. To fit models within these constraints, two broad strategies have emerge…