Dense Structural Compression of Transformers via Gauge-Correct Channel Removal

• URL: http://arxiv.org/abs/2609.07264v1
• Date: Mon, 07 Sep 2026 09:20:59 GMT
• Title: Dense Structural Compression of Transformers via Gauge-Correct Channel Removal
• Abstract summary: Inference energy per token drives the cost and carbon footprint of deployed transformers.<n>We develop a methodology to adapt structural complexity during training to maximize utility per unit compute.<n>Channel penalties drive entire tensor slices to zero to enable physical removal while preserving density and the network function.
• Score: 11.681900731333926
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Inference energy per token drives the cost and carbon footprint of deployed transformers. It is dominated by dense matrix products that incur fused multiply-accumulate (FMA) operations and memory traffic. To reduce these computations while retaining dense tensors for high GPU throughput, we develop a methodology from first principles to adapt structural complexity during training to maximize inference utility per unit compute. Channel penalties drive entire tensor slices to zero to enable physical removal while preserving density and the network function. The natural approach, penalizing the norm of operator components acting through each channel, is provably destabilized by gauge freedom. We resolve this pathology with GaugeLasso: additive symmetric group-lasso penalties that recover a monotone function of product-norms when the network converges to gauge balance. Our equilibrium analysis enables per-channel calibration to correctly suppress slices that under-perform in inference utility per unit compute. Under adaptive pressure, the network reorganizes into depth-dependent structural profiles that can be far smaller than the architecture required to learn the task. On polynomial long division over $\mathbb{F}_{31}$, compute compresses from 148 to 255 times with perfect accuracy. On character-level language modeling, compressed models outperform the hand-designed baseline at equal FMA. On masked autoencoding, a compression trial exposes which axes were over-provisioned and which saturated, guiding a better second design. Compaction also accelerates training monotonically as the model progresses. Post-hoc pruning with the same utility ranking cannot reach these structures, showing that sustained pressure is central to discovery of efficient models. Retraining a discovered architecture recovers baseline quality on our statistical tasks, but fails on our exact algorithmic task.

Related papers

• Scaling Adaptive Depth with Norm-Agnostic Residual Networks [9.219329760729629]
  NAG is a norm-agnostic residual architecture that separates magnitude from directional information in the residual stream.<n>NAG introduces only a negligible number of additional parameters and relies on simple operations that are easily kernel-fusible.<n>We show that this architecture outperforms baseline Transformers, with gains that increase substantially as depth grows.
  arXiv  Detail & Related papers  (2026-06-15T02:05:37Z)
• Elastic Spiking Transformers for Efficient Gesture Understanding [6.52225846892392]
  Spiking Neural Networks (SNNs) offer energy-efficient processing of event-based sensor data for healthcare applications.<n>We introduce the Elastic Spiking Transformer, a runtime-adaptive architecture that brings elasticity into the spiking paradigm.<n>We evaluate the approach on CIFAR10/100, CIFAR10-DVS, and the EHWGesture clinical gesture understanding dataset.
  arXiv  Detail & Related papers  (2026-05-04T12:35:52Z)
• Alternating Gradient Flow Utility: A Unified Metric for Structural Pruning and Dynamic Routing in Deep Networks [52.153950303594684]
  We propose a decoupled kinetic paradigm inspired by Alternating Gradient Flow (AGF)<n>AGF accurately captures the network's structural "kinetic utility"<n>We design a hybrid routing framework that decouples AGF-guided offline structural search from online execution via zero-cost physical priors.
  arXiv  Detail & Related papers  (2026-03-12T18:19:21Z)
• Subjective Depth and Timescale Transformers: Learning Where and When to Compute [15.164635408299304]
  We introduce Subjective Depth Transformers (SDT) and Subjective Timescale Transformers (STT)<n>SDT and STT leverage Bayesian surprise signals to dynamically route computation, learning where and when to compute within decoder-only TFs.<n>The proposed architectures establish a flexible framework for efficiency, reducing self-attention computation by 75% and KV-cache requirements by 50% within each compute skipping layer, setting a pathway for more efficient models.
  arXiv  Detail & Related papers  (2025-11-26T14:00:18Z)
• PT$^2$-LLM: Post-Training Ternarization for Large Language Models [52.4629647715623]
  Large Language Models (LLMs) have shown impressive capabilities across diverse tasks, but their large memory and compute demands hinder deployment.<n>We propose PT$2$-LLM, a post-training ternarization framework tailored for LLMs.<n>At its core is an Asymmetric Ternary Quantizer equipped with a two-stage refinement pipeline.
  arXiv  Detail & Related papers  (2025-09-27T03:01:48Z)
• Convolutional Neural Network Compression Based on Low-Rank Decomposition [3.3295360710329738]
  This paper proposes a model compression method that integrates Variational Bayesian Matrix Factorization. VBMF is employed to estimate the rank of the weight tensor at each layer. Experimental results show that for both high and low compression ratios, our compression model exhibits advanced performance.
  arXiv  Detail & Related papers  (2024-08-29T06:40:34Z)
• Post-Training Network Compression for 3D Medical Image Segmentation: Reducing Computational Efforts via Tucker Decomposition [2.6820435361998918]
  We propose a post-training Tucker factorization that enables the decomposition of pre-existing models to reduce computational requirements without impeding segmentation accuracy. Our approach reduced the floating-point operations (FLOPs) and memory required during inference, offering an adjustable trade-off between computational efficiency and segmentation quality.
  arXiv  Detail & Related papers  (2024-04-15T11:36:31Z)
• Adaptive Fourier Neural Operators: Efficient Token Mixers for Transformers [55.90468016961356]
  We propose an efficient token mixer that learns to mix in the Fourier domain. AFNO is based on a principled foundation of operator learning. It can handle a sequence size of 65k and outperforms other efficient self-attention mechanisms.
  arXiv  Detail & Related papers  (2021-11-24T05:44:31Z)
• LCS: Learning Compressible Subspaces for Adaptive Network Compression at Inference Time [57.52251547365967]
  We propose a method for training a "compressible subspace" of neural networks that contains a fine-grained spectrum of models. We present results for achieving arbitrarily fine-grained accuracy-efficiency trade-offs at inference time for structured and unstructured sparsity. Our algorithm extends to quantization at variable bit widths, achieving accuracy on par with individually trained networks.
  arXiv  Detail & Related papers  (2021-10-08T17:03:34Z)
• Efficient Micro-Structured Weight Unification and Pruning for Neural Network Compression [56.83861738731913]
  Deep Neural Network (DNN) models are essential for practical applications, especially for resource limited devices. Previous unstructured or structured weight pruning methods can hardly truly accelerate inference. We propose a generalized weight unification framework at a hardware compatible micro-structured level to achieve high amount of compression and acceleration.
  arXiv  Detail & Related papers  (2021-06-15T17:22:59Z)
• Learning Efficient GANs for Image Translation via Differentiable Masks and co-Attention Distillation [130.30465659190773]
  Generative Adversarial Networks (GANs) have been widely-used in image translation, but their high computation and storage costs impede the deployment on mobile devices. We introduce a novel GAN compression method, termed DMAD, by proposing a Differentiable Mask and a co-Attention Distillation. Experiments show DMAD can reduce the Multiply Accumulate Operations (MACs) of CycleGAN by 13x and that of Pix2Pix by 4x while retaining a comparable performance against the full model.
  arXiv  Detail & Related papers  (2020-11-17T02:39:19Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.