Seungcheol Park, Jeongin Bae, Beomseok Kwon, Minjun Kim, Byeongwook Kim, Se Jung Kwon, U Kang, Dongsoo Lee. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

How can we accelerate large language models (LLMs) without sacrificing accuracy? The slow inference speed of LLMs hinders us to benefit from their remarkable performance in diverse applications. This is mainly because numerous sublayers are stacked together in LLMs. Sublayer pruning compresses and expedites LLMs via removing unnecessary sublayers. However, existing sublayer pruning algorithms are limited in accuracy since they naively select sublayers to prune, overlooking the different characteristics of each sublayer. In this paper, we propose SPRINT (Sublayer Pruning with Latency and Tunability Information), an accurate sublayer pruning method for LLMs. SPRINT accurately selects a target sublayer to prune by considering 1) the amount of latency reduction after pruning and 2) the tunability of sublayers. SPRINT iteratively prunes redundant sublayers and swiftly tunes the parameters of remaining sublayers. Experiments show that SPRINT achieves the best accuracy-speedup trade-off, exhibiting up to 23.88%p higher accuracy on zero-shot commonsense reasoning benchmarks compared to existing pruning algorithms.

Given a pretrained encoder-based language model, how can we accurately compress it without retraining? Retraining-free structured pruning algorithms are crucial in pretrained language model compression due to their significantly reduced pruning cost and capability to prune large language models. However, existing retraining-free algorithms encounter severe accuracy degradation, as they fail to handle pruning errors, especially at high compression rates. In this paper, we propose KPrune (Knowledge-preserving pruning), an accurate retraining-free structured pruning algorithm for pretrained encoder-based language models. KPrune focuses on preserving the useful knowledge of the pretrained model to minimize pruning errors through a carefully designed iterative pruning process composed of knowledge measurement, knowledge-preserving mask search, and knowledge-preserving weight-tuning. As a result, KPrune shows significant accuracy improvements up to 58.02%p higher F1 score compared to existing retraining-free pruning algorithms under a high compression rate of 80% on the SQuAD benchmark without any retraining process.

Given 3D LiDAR point clouds, how can we segment them fast and accurately? Fast and accurate segmentation of 3D LiDAR points is an important issue in mobile robotics with various applications in classification, tracking, SLAM, etc. Despite its importance, existing methods do not provide both speed and accuracy; in particular, methods performing segmentation in the 3D domain are too slow, disabling its use in real-time processing. In this paper, we propose Curved-Voxel Clustering (CVC), a fast and accurate method for segmenting 3D LiDAR point clouds utilizing LiDAR-optimized curved-voxel. CVC attains fine discriminations by considering three important aspects for clustering 3D LiDAR points: distance from the sensor, directional resolutions, and rarity of points. CVC succeeds in providing real-time performance by carefully managing curved-voxels with a hash table. Especially, CVC works well on sparse 3D point clouds. Through experiments, we show that our method is up to 1.7× faster and 30% more accurate than other segmentation methods. CVC enables real-time segmentation with more than 20 runs in a second.