The paper introduces the "Compression Trinity," a unified framework that jointly applies sparsity, quantization, and low‑rank approximations to compress large language models. It presents several methods—MKOR, SLoPe, OPTIMA, PATCH, and SLiM—that leverage these three pillars to accelerate training, reduce memory bandwidth, and recover accuracy, achieving significant speedups and accuracy gains over existing techniques. The results demonstrate that combining all three compression strategies is essential for efficient, scalable, high‑performance LLM deployment.
By Mohammad Mozaffari
Prohibitive computational and environmental costs impede the scalable deployment of Large Language Models (LLMs). Traditional compression techniques (sparsity, quantization, low-rank approximations) a...
SHIFT-LLM is a training‑free post‑pruning correction framework that inserts a Linear Residual Adapter (LRA) at each depth‑pruned site in large language models. Each LRA preserves the original residual identity while adding a lightweight affine correction calibrated via closed‑form least‑squares regression on a small held‑out set, thereby approximating the hidden state that would have been produced by the removed block. Experiments across multiple model families and benchmarks show that SHIFT‑LLM consistently recovers accuracy lost to depth pruning, achieving gains up to +15.7 points on Llama‑3.1‑8B‑Instruct with only a few hundred calibration samples and no gradient computation.
By Ali Bahri, Hang Li, Hongliang Li, Zhitang Chen
SparseDecoding is a decoding‑aware pruning framework that improves large language model inference by aligning pruning with the activations seen during actual autoregressive generation. It constructs calibration matrices from dense‑model activations (excluding prefill) and introduces an optimized N:M sparse matrix‑vector kernel with bitmask indexing to accelerate decoding. Experiments on models such as Llama‑3.1‑8B, Llama‑3.3‑70B, and Qwen3‑14B/32B show consistent performance gains over standard fixed‑text calibration and up to 1.48× wall‑clock speedup on A100 GPUs.
By Qitong Wang, Xinwei Niu, Mingluo Su, Shanwei Zhao, Shiai Zhu, Huan Wang
arXiv:2607. 25504v1 Announce Type: cross Abstract: Fine-grained weight pruning and activation sparsification have emerged as effective approaches for reducing the compute and memory cost of inference for Transformer models.
By Bowen Wang, Chi Zhang, Diyou Shen, Renzo Andri, Navaneeth Kunhi Purayil, Luca Benini
arXiv:2601. 16991v3 Announce Type: replace-cross Abstract: Adapting large pre-trained language models to downstream tasks often entails fine-tuning millions of parameters or deploying costly dense weight updates, which hinders their use in resource-constrained environments.
By Longteng Zhang, Sen Wu, Shuai Hou, Zhengyu Qing, Zhuo Zheng, Danning Ke, Qihong Lin, Qiang Wang, Shaohuai Shi, Xiaowen Chu
arXiv:2607. 08786v1 Announce Type: cross Abstract: With the growing deployment of large language models (LLMs), LLM inference cost has become a key challenge.
By Tao Lu, Haoyu Wang, Zonghui Wang, Keshen Xiang, Jiaheng Zhang, Wenzhi Chen
Modern transformers pair impressive capabilities with substantial memory and compute demands. Low-rank weight factorization reduces both while keeping the matrices dense, and thus efficient on standar...
arXiv:2411. 09816v5 Announce Type: replace Abstract: Large neural networks achieve state-of-the-art performance on many tasks, yet their sheer size hinders deployment on resource-constrained devices.
By Cem \"Uy\"uk, Mike Lasby, Mohamed Yassin, Utku Evci, Yani Ioannou
arXiv:2606. 10445v1 Announce Type: new Abstract: Semi-structured 2:4 sparsity is widely supported by modern accelerators, providing up to a 2x theoretical speedup.
By Jaeseong Lee, Seung-won Hwang, Samyam Rajbhandari
arXiv:2602. 00161v2 Announce Type: replace-cross Abstract: In this paper, we formulate the compression of large language models (LLMs) by optimally deleting transformer blocks (``block removal'') as a constrained binary optimization (CBO) problem that can be mapped to a physical system (Ising glass), whose energies are a strong proxy for downstream model performance.
By David Jansen, Roman Rausch, Ali Hashemi, David Montero, Rom\'an Or\'us
MoARa introduces a module-aware rank allocation strategy and a block-wise magnitude-direction decomposition to improve low-rank gradient projection for large language model pre‑training. By profiling Transformer modules and tailoring projection ranks, it reduces the number of steps and wall‑clock time needed to reach target perplexity. Experiments on Llama, Qwen, and DeepSeek models show up to 41.7% fewer steps and 37.1% less training time with minimal memory overhead.
By Keunyoung Kim, Nojun Kwak