Variance Is Not Importance: Structural Analysis of Transformer Compressibility Across Model Scales

• URL: http://arxiv.org/abs/2604.20682v1
• Date: Wed, 22 Apr 2026 15:31:46 GMT
• Title: Variance Is Not Importance: Structural Analysis of Transformer Compressibility Across Model Scales
• Abstract summary: spectral compression, block-level function replacement, rotation-based quantization, activation geometry, and adaptive early exit are studied.<n>We identify five structural properties relevant to compression.
• Score: 0.0
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: We present a systematic empirical study of transformer compression through over 40 experiments on GPT-2 (124M parameters) and Mistral 7B (7.24B parameters). Our analysis covers spectral compression, block-level function replacement, rotation-based quantization, activation geometry, and adaptive early exit. We identify five structural properties relevant to compression. (1) Variance is not importance: high-variance activation directions are approximately 96 percent uncorrelated with predictive directions (measured via CCA), and projecting onto these subspaces preserves over 90 percent of variance while degrading perplexity. (2) Block linearity is conditional: transformer blocks are approximately linear (R^2 ~ 0.95 on GPT-2, 0.93 on Mistral block 31) only under the correct upstream distribution; modifying earlier blocks induces distribution shift that degrades downstream approximations. (3) The reconstruction wall: approaches that factor weights into quantized components amplify errors through cross-terms, making direct quantization strictly superior. (4) Linearity increases with depth: Mistral 7B exhibits a progression from R^2 = 0.17 (block 0) to R^2 = 0.93 (block 31), indicating a division between nonlinear feature construction and linear refinement. (5) Approximately 30 percent of tokens are computationally easy, confirmed via exit heads and KL divergence sensitivity. We demonstrate that single-block linear replacement achieves 34x compression with a 1.71 perplexity increase on the final block of Mistral 7B, while multi-block replacement fails due to residual error accumulation and distribution shift. These findings suggest fundamental limits to static post-training compression and motivate adaptive, per-token computation as a more effective direction.

Related papers

• Feature Priming in Online Linear Regression: Sparse-Regret Lower Bounds and a Tight Univariate Rate [47.36630149538994]
  In high-dimensional online prediction, the best predictor may depend on only a few features, so regret should scale with sparsity rather than the ambient dimension.<n>Warmuth and Amid asked at COLT 2023 whether any of three such rules admits a competitive online regret guarantee.<n>Our analysis identifies a common obstruction: cheap nuisance causes the refit to underweight the truly predictive coordinate.
  arXiv  Detail & Related papers  (2026-08-18T09:32:03Z)
• Orientation, not magnitude: the causal structure of task-vector interference in merged language models [0.0]
  We show that task arithmetic works until it doesn't, and the field diagnoses why with magnitudes.<n>An exact decomposition of the layerwise flux shows it is dominated by amplifying transport of the existing cross-term.<n>The striking +-15% "universality" of naive bfloat16 generation turns out to be quantization roughness.
  arXiv  Detail & Related papers  (2026-08-12T08:40:24Z)
• KroQuant: Kronecker-Structured Block Transforms for Efficient Post-Training Quantization of Diffusion Transformers [42.11461757540461]
  Post-training quantization of diffusion transformers (DiTs) to W4A4 severely degrades output quality.<n>We propose KroQuant, a PTQ method that applies a learned Kronecker-structured invertible transform to each 32-element block of the activation.
  arXiv  Detail & Related papers  (2026-07-23T15:52:20Z)
• Non-Vacuous Certification of Transport MCMC via Oscillation-Controlled Normalizing Flows [3.8549131582427303]
  Transport MCMC trains a normalizing flow to precondition Metropolis--Hastings proposals.<n>We establish the first numerically non-vacuous, rigorous spectral-gap bound for such samplers.
  arXiv  Detail & Related papers  (2026-05-31T07:46:48Z)
• Gated Subspace Inference for Transformer Acceleration [0.0]
  A method is presented for accelerating inference in transformer language models.<n>The method decomposes each activation vector into a subspace component and a residual.<n>It computes the linear-layer output on the subspace component via a cached low-rank weight image at reduced memory bandwidth.
  arXiv  Detail & Related papers  (2026-05-04T19:48:40Z)
• Compressing Transformer Language Models via Matrix Product Operator Decomposition: A Case Study on PicoGPT [0.0]
  Transformer-based language models achieve strong performance across NLP tasks, but their quadratic parameter scaling with hidden dimension makes deployment on resource-constrained hardware expensive.<n>We study Matrix Product Operator decomposition as a principled compression method for transformers.<n>MPO factorises weight matrices into chains of low-rank cores, with approximation quality controlled by the bond dimension chi.
  arXiv  Detail & Related papers  (2026-03-30T14:57:47Z)
• SVG-EAR: Parameter-Free Linear Compensation for Sparse Video Generation via Error-aware Routing [77.91660464664615]
  Diffusion Transformers (DiTs) have become a leading backbone for video generation, yet their quadratic attention cost remains a major bottleneck.<n>In this paper, we show that the missing contributions can be recovered without training.<n>We introduce SVG-EAR, a parameter-free linear compensation branch that uses the centroid to approximate skipped blocks and recover their contributions.
  arXiv  Detail & Related papers  (2026-03-09T22:15:31Z)
• Information Fidelity in Tool-Using LLM Agents: A Martingale Analysis of the Model Context Protocol [69.11739400975445]
  We introduce the first theoretical framework for analyzing error accumulation in Model Context Protocol (MCP) agents.<n>We show that cumulative distortion exhibits linear growth and high-probability deviations bounded by $O(sqrtT)$.<n>Key findings include: semantic weighting reduces distortion by 80%, and periodic re-grounding approximately every 9 steps suffices for error control.
  arXiv  Detail & Related papers  (2026-02-10T21:08:53Z)
• Learning on the Manifold: Unlocking Standard Diffusion Transformers with Representation Encoders [48.68968421120471]
  We show that standard diffusion transformers fail to converge on representations directly.<n>We identify Geometric Interference as the root cause.<n>Our method RJF enables the standard DiT-B architecture to converge effectively, achieving an FID of 3.37.
  arXiv  Detail & Related papers  (2026-02-10T18:58:04Z)
• Block-Recurrent Dynamics in Vision Transformers [42.261020313952976]
  We argue that trained ViTs admit a block-recurrent depth structure such that the computation of the original $L$ blocks can be accurately rewritten using only $k ll L$ distinct blocks applied recurrently.<n>We train a Raptor model to recover $96%$ of DINOv2 ImageNet-1k linear probe accuracy in only 2 blocks at equivalent computational cost.
  arXiv  Detail & Related papers  (2025-12-23T00:18:23Z)
• Pay Attention Later: From Vector Space Diffusion to Linearithmic Spectral Phase-Locking [0.0]
  Standard Transformers suffer from a "Semantic Alignment Tax"<n>We introduce the Phase-Resonant Intelligent Spectral Model (PRISM)<n>PRISM encodes semantic identity as resonant frequencies in the complex domain (Cd) and replaces quadratic self-attention with linearithmic O(N log N) Gated Harmonic Convolutions.
  arXiv  Detail & Related papers  (2025-12-01T02:46:15Z)
• MLPMoE: Zero-Shot Architectural Metamorphosis of Dense LLM MLPs into Static Mixture-of-Experts [0.0]
  Large Language Models (LLMs) are predominantly deployed as dense transformers, where every parameter in every feed-forward block is activated for every token.<n>Recent upcycling methods such as MoEfication, CMoE, ToMoE, and MoORE reveal that much of the useful computation lives in sparse, semi-modular substructures inside dense feed-forward networks.<n>This paper introducesMoE (MLP-Experts), a training-free transformation that restructures the dense in transformer blocks into a static, high-cardinality mixture of experts.
  arXiv  Detail & Related papers  (2025-11-26T06:14:26Z)
• Change-of-Basis Pruning via Rotational Invariance [0.0]
  We introduce two-subspace activations (TSRAs) to allow CoB transformations to be merged into surrounding weights without incurring extra parameters.<n>This work is a proof-of-concept that a rotationally invariant design may offer a principled approach towards change-of-basis pruning.
  arXiv  Detail & Related papers  (2025-11-20T05:45:19Z)
• Temporal Zoom Networks: Distance Regression and Continuous Depth for Efficient Action Localization [6.908972852063454]
  Temporal action localization requires both precise boundary detection and computational efficiency.<n>We address this through two complementary innovations: Boundary Distance Regression (BDR) and Adaptive Temporal Refinement (ATR)<n>On THUMOS14, our method achieves 56.5% mAP@0.7 with 151G FLOPs, using 36% fewer FLOPs than ActionFormer++ (55.7% mAP@0.7 at 235G)
  arXiv  Detail & Related papers  (2025-11-06T00:41:54Z)
• Rao-Blackwell Gradient Estimators for Equivariant Denoising Diffusion [55.95767828747407]
  In domains such as molecular and protein generation, physical systems exhibit inherent symmetries that are critical to model.<n>We present a framework that reduces training variance and provides a provably lower-variance gradient estimator.<n>We also present a practical implementation of this estimator incorporating the loss and sampling procedure through a method we call Orbit Diffusion.
  arXiv  Detail & Related papers  (2025-02-14T03:26:57Z)
• Adaptive Multi-step Refinement Network for Robust Point Cloud Registration [82.64560249066734]
  Point Cloud Registration estimates the relative rigid transformation between two point clouds of the same scene.<n>We propose an adaptive multi-step refinement network that refines the registration quality at each step by leveraging the information from the preceding step.<n>Our method achieves state-of-the-art performance on both the 3DMatch/3DLoMatch and KITTI benchmarks.
  arXiv  Detail & Related papers  (2023-12-05T18:59:41Z)
• Pruning Redundant Mappings in Transformer Models via Spectral-Normalized Identity Prior [54.629850694790036]
  spectral-normalized identity priors (SNIP) is a structured pruning approach that penalizes an entire residual module in a Transformer model toward an identity mapping. We conduct experiments with BERT on 5 GLUE benchmark tasks to demonstrate that SNIP achieves effective pruning results while maintaining comparable performance.
  arXiv  Detail & Related papers  (2020-10-05T05:40:56Z)
• Analytic Signal Phase in $N-D$ by Linear Symmetry Tensor--fingerprint modeling [69.35569554213679]
  We show that the Analytic Signal phase, and its gradient have a hitherto unstudied discontinuity in $2-D $ and higher dimensions. This shortcoming can result in severe artifacts whereas the problem does not exist in $1-D $ signals. We suggest the use of Linear Symmetry phase, relying on more than one set of Gabor filters, but with a negligible computational add-on.
  arXiv  Detail & Related papers  (2020-05-16T21:17:26Z)

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.