COEC: Calibrated Orthogonal-Equivalence Compensation for Structured Pruning of Large Language Models

• URL: http://arxiv.org/abs/2608.21142v1
• Date: Fri, 21 Aug 2026 14:20:49 GMT
• Title: COEC: Calibrated Orthogonal-Equivalence Compensation for Structured Pruning of Large Language Models
• Abstract summary: We propose COEC (Calibrated Orthogonal-Equivalence Compensation) as a training-free compensation framework for structured pruning.<n>Experiments on the Llama-3, Llama-3.1, and Qwen2.5 model families show that COEC improves perplexity on every model and zero-shot accuracy in most settings.
• Score: 13.501861808796873
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Structured pruning reduces the size and inference cost of large language models (LLMs) by removing weight columns, but the resulting output error can degrade accuracy. Existing training-free compensation methods use an additive bias or a single orthogonal rotation on the output side of the retained weight. These corrections leave its input singular frame unchanged and therefore limit how the retained weight can adapt after column removal. We propose COEC (Calibrated Orthogonal-Equivalence Compensation), a training-free compensation framework that applies alternating left and right orthogonal rotations to the retained weight. The right rotation is optimized on a reduced Stiefel manifold, while singular values are rescaled using generalized cross-validation to select the regularization strength for each layer. COEC further tempers the calibration Gram matrix to reduce the dominance of high-energy activation directions and introduces an alignment penalty that preserves the geometric relation between adjacent attention projections.All components use second-order statistics from a small calibration set and require neither backpropagation through the LLM nor retraining of the model parameters. COEC is independent of the column pruning criterion and can be applied to multiple structured pruning methods. Experiments on the Llama-3, Llama-3.1, and Qwen2.5 model families across multiple structured sparsity levels show that COEC improves perplexity on every model and zero-shot accuracy in most settings over existing compensation methods, with larger gains at higher sparsity. These results show that post-pruning compensation can recover part of the performance lost to column removal.

Related papers

• SHIFT-LLM: Distribution Shift Correction in Depth-Pruned LLMs [18.494055578192015]
  We introduce SHIFT-LLM, a training-free post-pruning correction framework that inserts a Linear Residual Adapter at each pruning site.<n>Each LRA preserves the identity pathway of the original residual block and adds a lightweight affine residual correction.<n>Experiments on five model families, six layer-selection criteria, and seven zero-shot benchmarks show that SHIFT-LLM consistently recovers accuracy lost to depth pruning.
  arXiv  Detail & Related papers  (2026-08-25T19:01:58Z)
• KALE: Kernel Alignment with Loss Equilibration for Stable CLIP-DINOv2 Alignment at Web Scale [0.0]
  Kernel-based alignment of CLIP toward a vision centric teacher such as DINOv2 (KUEA) improves CLIP's visual representations while preserving text-encoder compatibility.<n>We introduce KALE, a loss-equilibration controller that tracks both losses and adaptively rescales the alignment weight toward a target ratio.<n>On a 3.3M-image CC12M subset, the aligned model preserves image-text retrieval and reproducibly improves SVHN linear probing.
  arXiv  Detail & Related papers  (2026-07-21T09:14:33Z)
• Improving Neural Network Training by Decoupling the Magnitude and Direction of Weight Vectors [110.10775872372047]
  We propose emphMagnitude-Direction-Direction (MD) Decoupling.<n>It factorizes each weight into a fixed-norm direction on a hypersphere and learnable per-row and per-column magnitude gains, updated at separate learning rates.<n>Across both Adam and Muon, MD Decoupling improves on well-tuned baselines, transfers the optimal LR across model width without retuning, and continues to help at scale on large Mixture-of-Experts models.
  arXiv  Detail & Related papers  (2026-06-24T15:40:26Z)
• HARP: Hadamard-Preconditioned Adaptive Rotation Processor for Extreme LLM Quantization [48.87937677659571]
  We introduce HARP (Hadamard-preconditioned Adaptive Rotation Processor), a learnable structured two-sided processor that replaces fixed Hadamard mixing.<n>Across 2-4 bit settings on models ranging from 1B to 70B parameters, HARP improves perplexity and zero-shot accuracy over fixed RHT.<n>HarP preserves deployment efficiency, reaching 128 tok/s versus 61 tok/s for FP16.
  arXiv  Detail & Related papers  (2026-05-28T12:24:15Z)
• Diffusion Controller: Framework, Algorithms and Parameterization [54.82539154511621]
  We present a unified control-theoretic view that casts reverse diffusion sampling as state-only control within (generalized) linearly-solvable Markov Decision Processes.<n>Under this framework, control acts by reweighting the pretrained reverse-time transition kernels, balancing terminal objectives against an $f$-divergence cost.<n>Experiments on Stable Diffusion v1.4 show consistent gains in preference-alignment win rates and improved quality-efficiency trade-offs.
  arXiv  Detail & Related papers  (2026-03-07T01:49:59Z)
• Correctness-Optimized Residual Activation Lens (CORAL): Transferrable and Calibration-Aware Inference-Time Steering [3.7758197704962835]
  We introduce CORAL, a regularized-time steering method that captures correctness signals from model internal activations using weight-decay probes.<n>CORAL consistently improves accuracy by 10% and expected calibration error (ECE) by 50% on average.<n>Our results support the hypothesis that distributed information in model internals can be extracted using regularized probes when individual neurons are insufficient.
  arXiv  Detail & Related papers  (2026-02-05T18:55:56Z)
• A Unified View of Attention and Residual Sinks: Outlier-Driven Rescaling is Essential for Transformer Training [86.64715217940274]
  Outliers function jointly with normalization.<n>Outliers serve more as rescale factors rather than contributors.<n>Outliers can be absorbed into learnable parameters or mitigated via explicit gated rescaling.
  arXiv  Detail & Related papers  (2026-01-30T13:29:45Z)
• RCPU: Rotation-Constrained Error Compensation for Structured Pruning of a Large Language Model [4.7618979083425215]
  We propose a rotation-constrained compensation method to address the errors introduced by structured pruning of large language models (LLMs)<n>LLMs are trained on massive datasets and accumulate rich semantic knowledge in their representation space.<n>In experiments, we apply the proposed method to LLaMA-7B and evaluate it on WikiText-2 and multiple language understanding benchmarks.
  arXiv  Detail & Related papers  (2025-10-09T04:54:09Z)
• 3-Dimensional CryoEM Pose Estimation and Shift Correction Pipeline [2.009945677846956]
  Accurate pose estimation and shift correction are key challenges in cryo-EM due to the very low SNR, which directly impacts the fidelity of 3D reconstructions.<n>We present an approach for pose estimation in cryo-EM that leverages multi-dimensional scaling (MDS) techniques in a robust manner to estimate the 3D rotation matrix of each particle from pairs of dihedral angles.
  arXiv  Detail & Related papers  (2025-07-20T11:46:17Z)
• UPCORE: Utility-Preserving Coreset Selection for Balanced Unlearning [57.081646768835704]
  User specifications or legal frameworks often require information to be removed from pretrained models, including large language models (LLMs)<n>This requires deleting or "forgetting" a set of data points from an already-trained model, which typically degrades its performance on other data points.<n>We propose UPCORE, a method-agnostic data selection framework for mitigating collateral damage during unlearning.
  arXiv  Detail & Related papers  (2025-02-20T22:51:10Z)
• RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models [53.571195477043496]
  We propose an algorithm named Rotated Straight-Through-Estimator (RoSTE)<n>RoSTE combines quantization-aware supervised fine-tuning (QA-SFT) with an adaptive rotation strategy to reduce activation outliers.<n>Our findings reveal that the prediction error is directly proportional to the quantization error of the converged weights, which can be effectively managed through an optimized rotation configuration.
  arXiv  Detail & Related papers  (2025-02-13T06:44:33Z)
• Semi-Supervised Coupled Thin-Plate Spline Model for Rotation Correction and Beyond [84.56978780892783]
  We propose CoupledTPS, which iteratively couples multiple TPS with limited control points into a more flexible and powerful transformation. In light of the laborious annotation cost, we develop a semi-supervised learning scheme to improve warping quality by exploiting unlabeled data. Experiments demonstrate the superiority and universality of CoupledTPS over the existing state-of-the-art solutions for rotation correction.
  arXiv  Detail & Related papers  (2024-01-24T13:03:28Z)
• Classifier Calibration with ROC-Regularized Isotonic Regression [0.0]
  We use isotonic regression to minimize the cross entropy on a calibration set via monotone transformations. IR acts as an adaptive binning procedure, which allows achieving a calibration error of zero, but leaves open the issue of the effect on performance. We show empirically that this general monotony criterion is effective in striking a balance between reducing cross entropy loss and avoiding overfitting of the calibration set.
  arXiv  Detail & Related papers  (2023-11-21T08:45:09Z)

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.