The paper introduces Conditional Program (CP), a compressed representation for large language model memory tables that reconstructs only the rows requested by an input. CP encodes each row with a short two‑bit code and a local scale, using a shared reconstruction program based on a Gaussian trellis. Experiments on three n‑gram memory families show CP reduces output KL divergence by 41–63% compared to 2‑bit round‑to‑nearest, and for DeepSeek it cuts table storage from 203 GB to 51 GB, enabling larger workloads on GPU.
By Donghoon Han, Minseop Jung, Dong Woo Kwak, SeungJae Lee, Junhoo Lee
The paper investigates reducing memory usage of n‑gram embedding tables in open‑weight language models by storing them in low precision. Experiments on nanochat, Llama, and GPT‑2 families show that four‑bit tables cut table memory by about 75% while retaining over 99% of the validation‑loss improvement. The study also demonstrates that re‑allocating the saved memory to larger tables can surpass the performance of full‑precision tables, and that similar savings are achievable in large models like DeepSeek‑V4.1‑Flash and Qwen3.8‑Flash‑Next with minimal impact on perplexity and downstream tasks.
By Stephen Chung, Wenyu Du
The paper introduces COUNT, a method that builds n‑gram embeddings without adding learned parameters by storing precomputed n‑gram counts on host memory. By weighting the language model’s own unembeddings with corpus frequencies, COUNT can be added as an extra input embedding during pretraining, improving validation loss equivalent to a 57% increase in compute for a 1B‑parameter model. Additionally, COUNT enables efficient GPU memory usage and allows domain‑specific n‑gram tables to be swapped in at inference time for training‑free adaptation.
By Ian Magnusson, Noah A. Smith, Pang Wei Koh
arXiv:2605. 08692v2 Announce Type: replace Abstract: Post-training weight-only quantization to 4 bits is widely used to reduce the memory and compute costs of large language model inference.
By Beshr IslamBouli, David Jin
arXiv:2610.27699v1 Announce Type: new
Abstract: Low-bit quantization reduces the memory footprint of large language models, but preserving accuracy with efficient training remains challenging. Scalar...
By Xinyu Lei, Hong Huang, Jiebin Zhang, Hongyu Li, Jinlong Song, Decheng Wu, Guanghua Yu, Jianchen Zhu, Sujian Li
arXiv:2607. 07144v1 Announce Type: new Abstract: The key-value (KV) cache dominates the memory cost of long-context autoregressive inference, and a growing body of work compresses it through quantization, eviction, or offloading.
By Vladimir Gusev
arXiv:2601. 07372v2 Announce Type: replace-cross Abstract: While Mixture-of-Experts (MoE) scales capacity via conditional computation, Transformers lack a native primitive for knowledge lookup, forcing them to inefficiently simulate retrieval through computation.
By Xin Cheng, Rui Tian, Wangding Zeng, Damai Dai, Qinyu Chen, Bingxuan Wang, Zhenda Xie, Kezhao Huang, Xingkai Yu, Chengqi Deng, Shangyan Zhou, Chenggang Zhao, Zhewen Hao, Yukun Li, Han Zhang, Zhengyan Zhang, Yixu Wei, M. Y Xu, Huishuai Zhang, Dongyan Zhao, Wenfeng Liang
OrBIT is a structure‑guided embedding compression framework that learns reusable local geometry from orbit dynamics to constrain a small set of shared codewords. It allocates a fixed coding budget based on a global reconstruction residual, allowing redundant overlapping charts to compensate local errors after gluing. Across four large language model embedding tables, OrBIT achieves 37.9× compression on GPT‑2 and over 23× on each 7B table compared to 16‑bit storage, while maintaining competitive rate‑distortion performance against existing quantization and low‑rank methods.
By Yunied Puig, Amit Kumar Jaiswal
arXiv:2608. 15602v1 Announce Type: cross Abstract: While binary quantization theoretically promises extreme compression and acceleration for Large Language Models (LLMs), existing research often overlooks the necessity of specialized hardware kernels, thus failing to unleash the full acceleration potential due to persistent reliance on expensive floating-point arithmetic or runtime dequantization overheads.
By Qingyao Yang, Runming Yang, He Xiao, Wendong Xu, Junyu Chen, Haobo Liu, Chenchen Ding, Ruihan Hu, Yik-Chung Wu, Ngai Wong
arXiv:2505.10202v2 Announce Type: replace
Abstract: Large Language Models (LLMs) have achieved remarkable success but face significant computational and memory challenges, particularly due to their e...
By Jintian Shao, Hongyi Huang, Jiayi Wu, YiMing Cheng, ZhiYu Wu, You Shan, MingKai Zheng
arXiv:2601. 21461v3 Announce Type: replace-cross Abstract: Modern sparse language models typically achieve sparsity through Mixture-of-Experts (MoE) layers, which dynamically route tokens to dense MLP "experts.
By Albert Tseng, Christopher De Sa
Decoder-only language models entangle long-term memory and reasoning in a single parameter set, making it difficult to scale memory capacity independently. Memory Decoder introduces a parametric long-term memory module but only studies it at a relatively small scale.