Training and serving Large Language Models (LLMs) relies heavily on parallelization and collective operations, which are frequently bottlenecked by network bandwidth. Lossless compression using e.g., Huffman codes can alleviate the issue, however, Huffman codes suffer from slow, bit-sequential decoding and high hardware complexity due to deep tree traversals. Universal codes e.g., Exponential-Golomb codes are faster to decode but do not exploit the symbol frequency distributions. To address these limitations, this paper introduces Quad Length Codes, a hybrid approach designed to balance compression efficiency with decoding speed. The coding scheme uses 3 prefix bits to divide the 256 symbols into 8 areas. Each area has a different code length and encodes a different number of symbols. The scheme uses a Look Up Table with 256 entries, significantly simplifying the hardware implementation compared to Huffman trees. The coding scheme can be adapted for different distributions. For the e4m3 data type, the scheme achieves a compressibility of 13.9% in comparison to 15.9% achieved by Huffman codes, but it significantly speeds up the decoding and simplifies the hardware complexity.
Figures & tables
Figure 1: Sorted Probability Mass Function (PMF) of FFN1 activation.
Figure 2: Huffman code lengths for each symbol.
Area
Area code
#Symbols
#Symbol bits
Code length
Symbol Range
1
000
8
3
6
0-7
2
001
8
3
6
8-15
3
010
8
3
6
16-23
4
011
8
3
6
24-31
5
100
8
3
6
32-39
6
101
16
4
7
40-55
Table 1: Quad length coding scheme.
Figure 3: Code lengths for Huffman codes and quad length codes.
Figure 4: Sorted Probability Mass Function (PMF) of FFN2 activation.
Figure 5: Huffman code lengths for each symbol.
Area
Area code
#Symbols
#Symbol bits
Code length
Symbol Range
1
000
2
1
4
0-1
2
001
8
3
6
2-9
3
010
8
3
6
10-17
4
011
8
3
6
18-25
5
100
8
3
6
26-33
6
101
32
5
8
34-65
Table 2: Quad length coding scheme.
Figure 6: Code lengths for Huffman codes and quad length codes.
Figure 7: Probability Mass Function (PMF) of FFN1 activation.
Large Language Models (LLMs) have achieved impressive performance across diverse domains but remain inefficient during the autoregressive decoding phase. Unlike the prefill stage, which employs compute-bound GEMM operations, decoding executes a sequence of small GEMV-like computations that are memory-bound and underutilize modern accelerators. Weight-only vector quantization (VQ) has emerged as an effective compression technique that clusters model weights into a shared codebook and replaces the original weight matrix with low-precision indices, enabling 2-bit-level weight compression. While this approach substantially reduces model size and memory bandwidth, it still suffers from two critical inefficiencies: the low utilization of GEMV computation and frequent memory conflicts during codebook lookups. This paper presents EVA, an efficient vector-quantization-based architecture that addresses both computational and memory bottlenecks in LLM decoding. EVA builds on a simple yet effective insight that combines input-codebook computation with conflict-free memory access. Instead of reconstructing quantized weights from indices, EVA directly performs dot products between input vectors and the weight codebook, transforming LLM decoding from GEMV to GEMM computation. It then performs structured lookups from an intermediate output buffer, eliminating memory bank conflicts. We further design a hardware-software co-optimized architecture specialized for LLM decoding while remaining compatible with conventional prefill execution. Evaluations show that EVA achieves up to 11.17× speedup and 7.17× higher energy efficiency compared with the SOTA lookup-based architecture, while preserving arithmetic precision after vector quantization. Our code is available at https://github.com/dbw6/Eva.git.
Model quantization has become essential for efficient large language model deployment, yet existing approaches involve clear trade-offs: methods such as GPTQ and AWQ achieve practical compression but are lossy, while lossless techniques preserve fidelity but typically do not accelerate inference. This paper explores the middle ground of statistically-lossless compression through three complementary notions of losslessness for quantized LLMs. First, task-lossless compression preserves zero-shot benchmark accuracy within natural sampling variance and remains achievable at aggressive bitwidths. Second, we formalize the stricter notion of distribution-lossless compression, requiring the quantized model's next-token distribution to be practically indistinguishable from the original, and propose the Expected Acceptance Rate (EAR), the maximum token-agreement probability under optimal coupling, as a directly interpretable fidelity metric (for example, EAR >= 0.99 indicates 99% agreement). Third, we prove a gamma-squared variance law showing that symmetric quantization inflates noise variance by gamma squared relative to asymmetric quantization, making asymmetry necessary for distribution-lossless fidelity but not for task-level preservation. Using SLQ, a layer-wise non-uniform method with asymmetric quantization and wide bitwidth search, we achieve task-lossless compression at well below 4 bits per parameter (as low as 3.3 bits depending on the model), distribution-lossless compression at 5 to 6 bits per parameter on average, and inference speedups of 1.7 to 3.6x relative to FP16 with optimized kernels. Source code is available at https://github.com/IST-DASLab/SLQ.
Large language models (LLMs) have shown strong potential for lossless data compression, but existing approaches are constrained by the high computational cost and low throughput of autoregressive decoding. We propose HyperZip, an efficient and scalable LLM-based compression framework that leverages diffusion-based LLMs (dLLMs) with Multi-Token Prediction (MTP) to accelerate LLM-based data compression processes. We identify a trade-off in diffusion-based compression, where increasing decoding throughput degrades the compression rate. To mitigate this trade-off, HyperZip employs a hypernetwork to generate data-specific updates from a context representation, adapting the dLLM to the target data without costly fine-tuning, resulting in a low compression rate and high throughput. Extensive experiments show that HyperZip achieves a superior trade-off between compression rate and speed compared with state-of-the-art baselines.